This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[PATCH] implement -Wrestrict for sprintf (PR 83688)
- From: Martin Sebor <msebor at gmail dot com>
- To: gcc-patches <gcc-patches at gcc dot gnu dot org>
- Date: Tue, 8 Oct 2019 17:51:42 -0600
- Subject: [PATCH] implement -Wrestrict for sprintf (PR 83688)
Attached is a resubmission of the -Wrestrict implementation for
the sprintf family of functions. The original patch was posted
in 2017 but never approved. This revision makes only a few minor
changes to the original code, mostly necessitated by rebasing on
the top of trunk.
The description from the original posting still applies today:
The enhancement works by first determining the base object (or
pointer) from the destination of the sprintf call, the constant
offset into the object (and subobject for struct members), and
its size. For each %s argument, it then computes the same data.
If it determines that overlap between the two is possible it
stores the data for the directive argument (including the size
of the argument) for later processing. After the whole call and
format string have been processed, the code then iterates over
the stored directives and their arguments and compares the size
and length of the argument against the function's overall output.
If they overlap it issues a warning.
The solution is pretty simple. The only details that might be
worth calling out are the addition of a few utility functions some
of which at first glance look like they could be replaced by calls
to existing utilities:
* array_elt_at_offset
* field_at_offset
* get_origin_and_offset
* alias_offset
Specifically, get_origin_and_offset looks like a dead ringer for
get_addr_base_and_unit_offset, except since the former is only
used for warnings it is less conservative. It also works with
SSA_NAMEs. This is also the function I expect to need to make
changes to (and fix bugs in). The rest of the functions are
general utilities that could perhaps be moved to tree.c at some
point when there is a use for them elsewhere (I have some work
in progress that might need at least one of them).
Another likely question worth addressing is why the sprintf
overlap detection isn't handled in gimple-ssa-warn-restrict.c.
There is an opportunity for code sharing between the two "passes"
but it will require some fairly intrusive changes to the latter.
Those feel out of scope for the initial solution.
Finally, because of new dependencies between existing classes in
the file, some code had to be moved around within it a bit. That
contributed to the size of the patch making the changes seem more
extensive than they really are.
Tested on x86_64-linux with Binutils/GDB and Glibc.
Martin
The original submission:
https://gcc.gnu.org/ml/gcc-patches/2017-07/msg00036.html
PR middle-end/83688 - check if buffers may overlap when copying strings using sprintf
gcc/ChangeLog:
PR middle-end/83688
* gimple-ssa-sprintf.c (format_result::alias_info): New struct.
(directive::argno): New member.
(format_result::aliases, format_result::alias_count): New data members.
(format_result::append_alias): New member function.
(fmtresult::dst_offset): New data member.
(pass_sprintf_length::call_info::dst_origin): New data member.
(pass_sprintf_length::call_info::dst_field, dst_offset): Same.
(char_type_p, array_elt_at_offset, field_at_offset): New functions.
(get_origin_and_offset): Same.
(format_string): Call it.
(format_directive): Call append_alias and set directive argument
number.
(maybe_warn_overlap): New function.
(pass_sprintf_length::compute_format_length): Call it.
(pass_sprintf_length::handle_gimple_call): Initialize new members.
* gcc/tree-ssa-strlen.c (): Also enable when -Wrestrict is on.
gcc/testsuite/ChangeLog:
PR tree-optimization/35503
* gcc.dg/tree-ssa/builtin-sprintf-warn-23.c: New test.
diff --git a/gcc/gimple-ssa-sprintf.c b/gcc/gimple-ssa-sprintf.c
index b11d7989d5e..b47ed019615 100644
--- a/gcc/gimple-ssa-sprintf.c
+++ b/gcc/gimple-ssa-sprintf.c
@@ -86,6 +86,7 @@ along with GCC; see the file COPYING3. If not see
#include "alloc-pool.h"
#include "vr-values.h"
#include "tree-ssa-strlen.h"
+#include "tree-dfa.h"
/* The likely worst case value of MB_LEN_MAX for the target, large enough
for UTF-8. Ideally, this would be obtained by a target hook if it were
@@ -106,9 +107,6 @@ namespace {
static int warn_level;
-struct call_info;
-struct format_result;
-
/* The minimum, maximum, likely, and unlikely maximum number of bytes
of output either a formatting function or an individual directive
can result in. */
@@ -133,80 +131,6 @@ struct result_range
unsigned HOST_WIDE_INT unlikely;
};
-/* The result of a call to a formatted function. */
-
-struct format_result
-{
- /* Range of characters written by the formatted function.
- Setting the minimum to HOST_WIDE_INT_MAX disables all
- length tracking for the remainder of the format string. */
- result_range range;
-
- /* True when the range above is obtained from known values of
- directive arguments, or bounds on the amount of output such
- as width and precision, and not the result of heuristics that
- depend on warning levels. It's used to issue stricter diagnostics
- in cases where strings of unknown lengths are bounded by the arrays
- they are determined to refer to. KNOWNRANGE must not be used for
- the return value optimization. */
- bool knownrange;
-
- /* True if no individual directive could fail or result in more than
- 4095 bytes of output (the total NUMBER_CHARS_{MIN,MAX} might be
- greater). Implementations are not required to handle directives
- that produce more than 4K bytes (leading to undefined behavior)
- and so when one is found it disables the return value optimization.
- Similarly, directives that can fail (such as wide character
- directives) disable the optimization. */
- bool posunder4k;
-
- /* True when a floating point directive has been seen in the format
- string. */
- bool floating;
-
- /* True when an intermediate result has caused a warning. Used to
- avoid issuing duplicate warnings while finishing the processing
- of a call. WARNED also disables the return value optimization. */
- bool warned;
-
- /* Preincrement the number of output characters by 1. */
- format_result& operator++ ()
- {
- return *this += 1;
- }
-
- /* Postincrement the number of output characters by 1. */
- format_result operator++ (int)
- {
- format_result prev (*this);
- *this += 1;
- return prev;
- }
-
- /* Increment the number of output characters by N. */
- format_result& operator+= (unsigned HOST_WIDE_INT);
-};
-
-format_result&
-format_result::operator+= (unsigned HOST_WIDE_INT n)
-{
- gcc_assert (n < HOST_WIDE_INT_MAX);
-
- if (range.min < HOST_WIDE_INT_MAX)
- range.min += n;
-
- if (range.max < HOST_WIDE_INT_MAX)
- range.max += n;
-
- if (range.likely < HOST_WIDE_INT_MAX)
- range.likely += n;
-
- if (range.unlikely < HOST_WIDE_INT_MAX)
- range.unlikely += n;
-
- return *this;
-}
-
/* Return the value of INT_MIN for the target. */
static inline HOST_WIDE_INT
@@ -461,7 +385,7 @@ public:
/* Construct a FMTRESULT object with all counters initialized
to MIN. KNOWNRANGE is set when MIN is valid. */
fmtresult (unsigned HOST_WIDE_INT min = HOST_WIDE_INT_MAX)
- : argmin (), argmax (), nonstr (),
+ : argmin (), argmax (), dst_offset (HOST_WIDE_INT_MIN), nonstr (),
knownrange (min < HOST_WIDE_INT_MAX),
mayfail (), nullp ()
{
@@ -475,7 +399,7 @@ public:
KNOWNRANGE is set when both MIN and MAX are valid. */
fmtresult (unsigned HOST_WIDE_INT min, unsigned HOST_WIDE_INT max,
unsigned HOST_WIDE_INT likely = HOST_WIDE_INT_MAX)
- : argmin (), argmax (), nonstr (),
+ : argmin (), argmax (), dst_offset (HOST_WIDE_INT_MIN), nonstr (),
knownrange (min < HOST_WIDE_INT_MAX && max < HOST_WIDE_INT_MAX),
mayfail (), nullp ()
{
@@ -498,6 +422,11 @@ public:
/* The range a directive's argument is in. */
tree argmin, argmax;
+ /* The starting offset into the destination of the formatted function
+ call of the %s argument that points into (aliases with) the same
+ destination array. */
+ HOST_WIDE_INT dst_offset;
+
/* The minimum and maximum number of bytes that a directive
results in on output for an argument in the range above. */
result_range range;
@@ -621,14 +550,29 @@ static bool
get_int_range (tree, HOST_WIDE_INT *, HOST_WIDE_INT *, bool, HOST_WIDE_INT,
const vr_values *);
+struct call_info;
+
/* Description of a format directive. A directive is either a plain
string or a conversion specification that starts with '%'. */
struct directive
{
+ directive (const call_info *inf, unsigned dno)
+ : info (inf), dirno (dno), argno (), beg (), len (), flags (),
+ width (), prec (), modifier (), specifier (), arg (), fmtfunc ()
+ { }
+
+ /* Reference to the info structure describing the call that this
+ directive is a part of. */
+ const call_info *info;
+
/* The 1-based directive number (for debugging). */
unsigned dirno;
+ /* The zero-based argument number of the directive's argument ARG in
+ the function's argument list. */
+ unsigned argno;
+
/* The first character of the directive and its length. */
const char *beg;
size_t len;
@@ -723,6 +667,130 @@ struct directive
}
};
+/* The result of a call to a formatted function. */
+
+struct format_result
+{
+ format_result ()
+ : range (), aliases (), alias_count (), knownrange (), posunder4k (),
+ floating (), warned () { /* No-op. */ }
+
+ ~format_result ()
+ {
+ XDELETEVEC (aliases);
+ }
+
+ /* Range of characters written by the formatted function.
+ Setting the minimum to HOST_WIDE_INT_MAX disables all
+ length tracking for the remainder of the format string. */
+ result_range range;
+
+ struct alias_info
+ {
+ directive dir; /* The directive that aliases the destination. */
+ HOST_WIDE_INT offset; /* The offset at which it aliases it. */
+ result_range range; /* The raw result of the directive. */
+ };
+
+ /* An array of directives whose pointer argument aliases a part
+ of the destination object of the formatted function. */
+ alias_info *aliases;
+ unsigned alias_count;
+
+ /* True when the range above is obtained from known values of
+ directive arguments, or bounds on the amount of output such
+ as width and precision, and not the result of heuristics that
+ depend on warning levels. It's used to issue stricter diagnostics
+ in cases where strings of unknown lengths are bounded by the arrays
+ they are determined to refer to. KNOWNRANGE must not be used for
+ the return value optimization. */
+ bool knownrange;
+
+ /* True if no individual directive could fail or result in more than
+ 4095 bytes of output (the total NUMBER_CHARS_{MIN,MAX} might be
+ greater). Implementations are not required to handle directives
+ that produce more than 4K bytes (leading to undefined behavior)
+ and so when one is found it disables the return value optimization.
+ Similarly, directives that can fail (such as wide character
+ directives) disable the optimization. */
+ bool posunder4k;
+
+ /* True when a floating point directive has been seen in the format
+ string. */
+ bool floating;
+
+ /* True when an intermediate result has caused a warning. Used to
+ avoid issuing duplicate warnings while finishing the processing
+ of a call. WARNED also disables the return value optimization. */
+ bool warned;
+
+ /* Preincrement the number of output characters by 1. */
+ format_result& operator++ ()
+ {
+ return *this += 1;
+ }
+
+ /* Postincrement the number of output characters by 1. */
+ format_result operator++ (int)
+ {
+ format_result prev (*this);
+ *this += 1;
+ return prev;
+ }
+
+ /* Increment the number of output characters by N. */
+ format_result& operator+= (unsigned HOST_WIDE_INT);
+
+ /* Add a directive to the sequence of those with potentially aliasing
+ arguments. */
+ void append_alias (const directive &, HOST_WIDE_INT, const result_range &);
+
+private:
+ /* Not copyable or assignable. */
+ format_result (format_result&);
+ void operator= (format_result&);
+};
+
+format_result&
+format_result::operator+= (unsigned HOST_WIDE_INT n)
+{
+ gcc_assert (n < HOST_WIDE_INT_MAX);
+
+ if (range.min < HOST_WIDE_INT_MAX)
+ range.min += n;
+
+ if (range.max < HOST_WIDE_INT_MAX)
+ range.max += n;
+
+ if (range.likely < HOST_WIDE_INT_MAX)
+ range.likely += n;
+
+ if (range.unlikely < HOST_WIDE_INT_MAX)
+ range.unlikely += n;
+
+ return *this;
+}
+
+void
+format_result::append_alias (const directive &d, HOST_WIDE_INT off,
+ const result_range &resrng)
+{
+ unsigned cnt = alias_count + 1;
+ alias_info *ar = XNEWVEC (alias_info, cnt);
+
+ for (unsigned i = 0; i != alias_count; ++i)
+ ar[i] = aliases[i];
+
+ ar[alias_count].dir = d;
+ ar[alias_count].offset = off;
+ ar[alias_count].range = resrng;
+
+ XDELETEVEC (aliases);
+
+ alias_count = cnt;
+ aliases = ar;
+}
+
/* Return the logarithm of X in BASE. */
static int
@@ -804,39 +872,6 @@ tree_digits (tree x, int base, HOST_WIDE_INT prec, bool plus, bool prefix)
return res;
}
-/* Given the formatting result described by RES and NAVAIL, the number
- of available in the destination, return the range of bytes remaining
- in the destination. */
-
-static inline result_range
-bytes_remaining (unsigned HOST_WIDE_INT navail, const format_result &res)
-{
- result_range range;
-
- if (HOST_WIDE_INT_MAX <= navail)
- {
- range.min = range.max = range.likely = range.unlikely = navail;
- return range;
- }
-
- /* The lower bound of the available range is the available size
- minus the maximum output size, and the upper bound is the size
- minus the minimum. */
- range.max = res.range.min < navail ? navail - res.range.min : 0;
-
- range.likely = res.range.likely < navail ? navail - res.range.likely : 0;
-
- if (res.range.max < HOST_WIDE_INT_MAX)
- range.min = res.range.max < navail ? navail - res.range.max : 0;
- else
- range.min = range.likely;
-
- range.unlikely = (res.range.unlikely < navail
- ? navail - res.range.unlikely : 0);
-
- return range;
-}
-
/* Description of a call to a formatted function. */
struct call_info
@@ -850,6 +885,18 @@ struct call_info
/* Called built-in function code. */
built_in_function fncode;
+ /* The "origin" of the destination pointer argument, which is either
+ the DECL of the destination buffer being written into or a pointer
+ that points to it, plus some offset. */
+ tree dst_origin;
+
+ /* For a destination pointing to a struct array member, the offset of
+ the member. */
+ HOST_WIDE_INT dst_field;
+
+ /* The offset into the destination buffer. */
+ HOST_WIDE_INT dst_offset;
+
/* Format argument and format string extracted from it. */
tree format;
const char *fmtstr;
@@ -2143,6 +2190,249 @@ format_character (const directive &dir, tree arg, const vr_values *vr_values)
return res.adjust_for_width_or_precision (dir.width);
}
+/* Return true if TYPE is one of the three narrow character types. */
+static inline bool
+char_type_p (tree type)
+{
+ return (type == char_type_node
+ || type == signed_char_type_node
+ || type == unsigned_char_type_node);
+}
+
+/* Determine the offset *INDEX of the first byte of an array element of
+ TYPE (possibly recursively) into which the byte offset OFF points.
+ On success set *INDEX to the offset of the first byte and return type.
+ Otherwise, if no such element can be found, return null. */
+
+static tree
+array_elt_at_offset (tree type, HOST_WIDE_INT off, HOST_WIDE_INT *index)
+{
+ gcc_assert (TREE_CODE (type) == ARRAY_TYPE);
+
+ tree eltype = type;
+ while (TREE_CODE (TREE_TYPE (eltype)) == ARRAY_TYPE)
+ eltype = TREE_TYPE (eltype);
+
+ if (!char_type_p (TREE_TYPE (eltype)))
+ eltype = TREE_TYPE (eltype);
+
+ if (eltype == type)
+ {
+ *index = 0;
+ return type;
+ }
+
+ HOST_WIDE_INT typsz = int_size_in_bytes (type);
+ HOST_WIDE_INT eltsz = int_size_in_bytes (eltype);
+ if (off < typsz * eltsz)
+ {
+ *index = (off / eltsz) * eltsz;
+ return TREE_CODE (eltype) == ARRAY_TYPE ? TREE_TYPE (eltype) : eltype;
+ }
+
+ return NULL_TREE;
+}
+
+/* Determine the offset *INDEX of the first byte of a struct member of TYPE
+ (possibly recursively) into which the byte offset OFF points. On success
+ set *INDEX to the offset of the first byte and return true. Otherwise,
+ if no such member can be found, return false. */
+
+static bool
+field_at_offset (tree type, HOST_WIDE_INT off, HOST_WIDE_INT *index)
+{
+ gcc_assert (RECORD_OR_UNION_TYPE_P (type));
+
+ for (tree fld = TYPE_FIELDS (type); fld; fld = TREE_CHAIN (fld))
+ {
+ if (TREE_CODE (fld) != FIELD_DECL || DECL_ARTIFICIAL (fld))
+ continue;
+
+ tree fldtype = TREE_TYPE (fld);
+ HOST_WIDE_INT fldoff = int_byte_position (fld);
+
+ /* If the size is not available the field is a flexible array
+ member. Treat this case as success. */
+ tree typesize = TYPE_SIZE_UNIT (fldtype);
+ HOST_WIDE_INT fldsize = (tree_fits_uhwi_p (typesize)
+ ? tree_to_uhwi (typesize)
+ : off);
+
+ if (fldoff + fldsize < off)
+ continue;
+
+ if (TREE_CODE (fldtype) == ARRAY_TYPE)
+ {
+ HOST_WIDE_INT idx = 0;
+ if (tree ft = array_elt_at_offset (fldtype, off, &idx))
+ fldtype = ft;
+ else
+ break;
+
+ *index += idx;
+ fldoff -= idx;
+ off -= idx;
+ }
+
+ if (RECORD_OR_UNION_TYPE_P (fldtype))
+ {
+ *index += fldoff;
+ return field_at_offset (fldtype, off - fldoff, index);
+ }
+
+ *index += fldoff;
+ return true;
+ }
+
+ return false;
+}
+
+/* For an expression X of pointer type, recursively try to find the same
+ origin (object or pointer) as Y it references and return such an X.
+ When X refers to a struct member, set *FLDOFF to the offset of the
+ member from the beginning of the "most derived" object. */
+
+static tree
+get_origin_and_offset (tree x, HOST_WIDE_INT *fldoff, HOST_WIDE_INT *off)
+{
+ if (!x)
+ return NULL_TREE;
+
+ switch (TREE_CODE (x))
+ {
+ case ADDR_EXPR:
+ x = TREE_OPERAND (x, 0);
+ return get_origin_and_offset (x, fldoff, off);
+
+ case ARRAY_REF:
+ {
+ tree offset = TREE_OPERAND (x, 1);
+ HOST_WIDE_INT idx = (tree_fits_uhwi_p (offset)
+ ? tree_to_uhwi (offset) : HOST_WIDE_INT_MAX);
+
+ tree eltype = TREE_TYPE (x);
+ if (TREE_CODE (eltype) == INTEGER_TYPE)
+ {
+ if (off)
+ *off = idx;
+ }
+ else if (idx < HOST_WIDE_INT_MAX)
+ *fldoff += idx * int_size_in_bytes (eltype);
+ else
+ *fldoff = idx;
+
+ x = TREE_OPERAND (x, 0);
+ return get_origin_and_offset (x, fldoff, NULL);
+ }
+
+ case MEM_REF:
+ if (off)
+ {
+ tree offset = TREE_OPERAND (x, 1);
+ *off = (tree_fits_uhwi_p (offset)
+ ? tree_to_uhwi (offset) : HOST_WIDE_INT_MAX);
+ }
+
+ x = TREE_OPERAND (x, 0);
+
+ if (off)
+ {
+ tree xtype = TREE_TYPE (TREE_TYPE (x));
+
+ /* The byte offset of the most basic struct member the byte
+ offset *OFF corresponds to, or for a (multidimensional)
+ array member, the byte offset of the array element. */
+ HOST_WIDE_INT index = 0;
+
+ if ((RECORD_OR_UNION_TYPE_P (xtype)
+ && field_at_offset (xtype, *off, &index))
+ || (TREE_CODE (xtype) == ARRAY_TYPE
+ && TREE_CODE (TREE_TYPE (xtype)) == ARRAY_TYPE
+ && array_elt_at_offset (xtype, *off, &index)))
+ {
+ *fldoff += index;
+ *off -= index;
+ fldoff = NULL;
+ }
+ }
+
+ return get_origin_and_offset (x, fldoff, NULL);
+
+ case COMPONENT_REF:
+ {
+ tree fld = TREE_OPERAND (x, 1);
+ *fldoff += int_byte_position (fld);
+
+ get_origin_and_offset (fld, fldoff, off);
+ x = TREE_OPERAND (x, 0);
+ return get_origin_and_offset (x, fldoff, off);
+ }
+
+ case SSA_NAME:
+ {
+ gimple *def = SSA_NAME_DEF_STMT (x);
+ if (is_gimple_assign (def))
+ {
+ tree_code code = gimple_assign_rhs_code (def);
+ if (code == ADDR_EXPR)
+ {
+ x = gimple_assign_rhs1 (def);
+ return get_origin_and_offset (x, fldoff, off);
+ }
+
+ if (code == POINTER_PLUS_EXPR)
+ {
+ tree offset = gimple_assign_rhs2 (def);
+ if (off)
+ *off = (tree_fits_uhwi_p (offset)
+ ? tree_to_uhwi (offset) : HOST_WIDE_INT_MAX);
+
+ x = gimple_assign_rhs1 (def);
+ return get_origin_and_offset (x, fldoff, NULL);
+ }
+ else if (code == VAR_DECL)
+ {
+ x = gimple_assign_rhs1 (def);
+ return get_origin_and_offset (x, fldoff, off);
+ }
+ }
+ else if (gimple_nop_p (def) && SSA_NAME_VAR (x))
+ x = SSA_NAME_VAR (x);
+ }
+
+ default:
+ break;
+ }
+
+ return x;
+}
+
+/* If ARG refers to the same (sub)object or array element as described
+ by DST and DST_FLD, return the byte offset into the struct member or
+ array element referenced by ARG. Otherwise return HOST_WIDE_INT_MIN
+ to indicate that ARG and DST do not refer to the same object. */
+
+static HOST_WIDE_INT
+alias_offset (tree arg, tree dst, HOST_WIDE_INT dst_fld)
+{
+ /* See if the argument refers to the same base object as the destination
+ of the formatted function call, and if so, try to determine if they
+ can alias. */
+ if (!arg || !dst || !ptr_derefs_may_alias_p (arg, dst))
+ return HOST_WIDE_INT_MIN;
+
+ /* The two arguments may refer to the same object. If they both refer
+ to a struct member, see if the members are one and the same. */
+ HOST_WIDE_INT arg_off = 0, arg_fld = 0;
+
+ tree arg_orig = get_origin_and_offset (arg, &arg_fld, &arg_off);
+
+ if (arg_orig == dst && arg_fld == dst_fld)
+ return arg_off;
+
+ return HOST_WIDE_INT_MIN;
+}
+
/* Return the minimum and maximum number of characters formatted
by the '%s' format directive and its wide character form for
the argument ARG. ARG can be null (for functions such as
@@ -2153,6 +2443,17 @@ format_string (const directive &dir, tree arg, const vr_values *vr_values)
{
fmtresult res;
+ if (warn_restrict)
+ {
+ /* See if ARG might alias the destination of the call with
+ DST_ORIGIN and DST_FIELD. If so, store the starting offset
+ so that the overlap can be determined for certain later,
+ when the amount of output of the call (including subsequent
+ directives) has been computed. Othewrwise, store HWI_MIN. */
+ res.dst_offset = alias_offset (arg, dir.info->dst_origin,
+ dir.info->dst_field);
+ }
+
/* Compute the range the argument's length can be in. */
int count_by = 1;
if (dir.specifier == 'S' || dir.modifier == FMT_LEN_l)
@@ -2163,7 +2464,7 @@ format_string (const directive &dir, tree arg, const vr_values *vr_values)
/* Now that we have a suitable node, get the number of
bytes it occupies. */
- count_by = int_size_in_bytes (node);
+ count_by = int_size_in_bytes (node);
gcc_checking_assert (count_by == 2 || count_by == 4);
}
@@ -2695,6 +2996,39 @@ maybe_warn (substring_loc &dirloc, location_t argloc,
res.min, avail_range.min, avail_range.max);
}
+/* Given the formatting result described by RES and NAVAIL, the number
+ of available in the destination, return the range of bytes remaining
+ in the destination. */
+
+static inline result_range
+bytes_remaining (unsigned HOST_WIDE_INT navail, const format_result &res)
+{
+ result_range range;
+
+ if (HOST_WIDE_INT_MAX <= navail)
+ {
+ range.min = range.max = range.likely = range.unlikely = navail;
+ return range;
+ }
+
+ /* The lower bound of the available range is the available size
+ minus the maximum output size, and the upper bound is the size
+ minus the minimum. */
+ range.max = res.range.min < navail ? navail - res.range.min : 0;
+
+ range.likely = res.range.likely < navail ? navail - res.range.likely : 0;
+
+ if (res.range.max < HOST_WIDE_INT_MAX)
+ range.min = res.range.max < navail ? navail - res.range.max : 0;
+ else
+ range.min = range.likely;
+
+ range.unlikely = (res.range.unlikely < navail
+ ? navail - res.range.unlikely : 0);
+
+ return range;
+}
+
/* Compute the length of the output resulting from the directive DIR
in a call described by INFO and update the overall result of the call
in *RES. Return true if the directive has been handled. */
@@ -2793,6 +3127,12 @@ format_directive (const call_info &info,
NUL that's appended after the format string has been processed. */
result_range avail_range = bytes_remaining (info.objsize, *res);
+ /* If the argument aliases a part of the destination of the formatted
+ call at offset FMTRES.DST_OFFSET append the directive and its result
+ to the set of aliases for later processing. */
+ if (fmtres.dst_offset != HOST_WIDE_INT_MIN)
+ res->append_alias (dir, fmtres.dst_offset, fmtres.range);
+
bool warned = res->warned;
if (!warned)
@@ -3078,6 +3418,10 @@ parse_directive (call_info &info,
return len - !*str;
}
+ /* Set the directive argument's number to correspond to its position
+ in the formatted function call's argument list. */
+ dir.argno = *argno;
+
const char *pf = pcnt + 1;
/* POSIX numbered argument index or zero when none. */
@@ -3482,6 +3826,136 @@ parse_directive (call_info &info,
return dir.len;
}
+/* Diagnose overlap between destination and %s directive arguments. */
+
+static void
+maybe_warn_overlap (call_info &info, format_result *res)
+{
+ /* Two vectors of 1-based indices corresponding to either certainly
+ or possibly aliasing arguments. */
+ auto_vec<int, 16> aliasarg[2];
+
+ /* Go through the array of potentially aliasing directives and collect
+ argument numbers of those that do or may overlap the destination
+ object given the full result. */
+ for (unsigned i = 0; i != res->alias_count; ++i)
+ {
+ const format_result::alias_info &alias = res->aliases[i];
+
+ enum { Possible = -1, None = 0, Certain = 1 } overlap = None;
+
+ /* If the precision is zero there is no overlap. (This only
+ considers %s directives and ignores %n.) */
+ if (alias.dir.prec[0] == 0 && alias.dir.prec[1] == 0)
+ continue;
+
+ if (alias.offset == HOST_WIDE_INT_MAX
+ || info.dst_offset == HOST_WIDE_INT_MAX)
+ overlap = Possible;
+ else if (alias.offset == info.dst_offset)
+ overlap = alias.dir.prec[0] == 0 ? Possible : Certain;
+ else
+ {
+ /* Dtermine overlap from the range of output and offsets
+ into the same destination as the source, and rule out
+ impossible overlap. */
+ unsigned HOST_WIDE_INT albeg = alias.offset;
+ unsigned HOST_WIDE_INT dstbeg = info.dst_offset;
+
+ unsigned HOST_WIDE_INT alend = albeg + alias.range.min;
+ unsigned HOST_WIDE_INT dstend = dstbeg + res->range.min - 1;
+
+ if ((albeg <= dstbeg && alend > dstbeg)
+ || (albeg >= dstbeg && albeg < dstend))
+ overlap = Certain;
+ else
+ {
+ alend = albeg + alias.range.max;
+ if (alend < albeg)
+ alend = HOST_WIDE_INT_M1U;
+
+ dstend = dstbeg + res->range.max - 1;
+ if (dstend < dstbeg)
+ dstend = HOST_WIDE_INT_M1U;
+
+ if ((albeg >= dstbeg && albeg <= dstend)
+ || (alend >= dstbeg && alend <= dstend))
+ overlap = Possible;
+ }
+ }
+
+ if (overlap == None)
+ continue;
+
+ /* Append the 1-based argument number. */
+ aliasarg[overlap != Certain].safe_push (alias.dir.argno + 1);
+
+ /* Disable any kind of optimization. */
+ res->range.unlikely = HOST_WIDE_INT_M1U;
+ }
+
+ tree arg0 = gimple_call_arg (info.callstmt, 0);
+ location_t loc = gimple_location (info.callstmt);
+
+ bool aliaswarn = false;
+
+ unsigned ncertain = aliasarg[0].length ();
+ unsigned npossible = aliasarg[1].length ();
+ if (ncertain && npossible)
+ {
+ /* If there are multiple arguments that overlap, some certainly
+ and some possibly, handle both sets in a single diagnostic. */
+ aliaswarn
+ = warning_at (loc, OPT_Wrestrict,
+ "%qE arguments %Z and maybe %Z overlap destination "
+ "object %qE",
+ info.func, aliasarg[0].address (), ncertain,
+ aliasarg[1].address (), npossible,
+ info.dst_origin);
+ }
+ else if (ncertain)
+ {
+ /* There is only one set of two or more arguments and they all
+ certainly overlap the destination. */
+ aliaswarn
+ = warning_n (loc, OPT_Wrestrict, ncertain,
+ "%qE argument %Z overlaps destination object %qE",
+ "%qE arguments %Z overlap destination object %qE",
+ info.func, aliasarg[0].address (), ncertain,
+ info.dst_origin);
+ }
+ else if (npossible)
+ {
+ /* There is only one set of two or more arguments and they all
+ may overlap (but need not). */
+ aliaswarn
+ = warning_n (loc, OPT_Wrestrict, npossible,
+ "%qE argument %Z may overlap destination object %qE",
+ "%qE arguments %Z may overlap destination object %qE",
+ info.func, aliasarg[1].address (), npossible,
+ info.dst_origin);
+ }
+
+ if (aliaswarn)
+ {
+ res->warned = true;
+
+ if (info.dst_origin != arg0)
+ {
+ /* If its location is different from the first argument of the call
+ point either at the destination object itself or at the expression
+ that was used to determine the overlap. */
+ loc = (DECL_P (info.dst_origin)
+ ? DECL_SOURCE_LOCATION (info.dst_origin)
+ : EXPR_LOCATION (info.dst_origin));
+ if (loc != UNKNOWN_LOCATION)
+ inform (loc,
+ "destination object referenced by %<restrict%>-qualified "
+ "argument 1 was declared here");
+ }
+ }
+}
+
/* Compute the length of the output resulting from the call to a formatted
output function described by INFO and store the result of the call in
*RES. Issue warnings for detected past the end writes. Return true
@@ -3521,10 +3995,11 @@ compute_format_length (call_info &info, format_result *res, const vr_values *vr)
/* The variadic argument counter. */
unsigned argno = info.argidx;
+ bool success = true;
+
for (const char *pf = info.fmtstr; ; ++dirno)
{
- directive dir = directive ();
- dir.dirno = dirno;
+ directive dir (&info, dirno);
size_t n = parse_directive (info, dir, res, pf, &argno, vr);
@@ -3532,18 +4007,23 @@ compute_format_length (call_info &info, format_result *res, const vr_values *vr)
if (!format_directive (info, res, dir, vr))
return false;
- /* Return success the directive is zero bytes long and it's
- the last think in the format string (i.e., it's the terminating
+ /* Return success when the directive is zero bytes long and it's
+ the last thing in the format string (i.e., it's the terminating
nul, which isn't really a directive but handling it as one makes
things simpler). */
if (!n)
- return *pf == '\0';
+ {
+ success = *pf == '\0';
+ break;
+ }
pf += n;
}
+ maybe_warn_overlap (info, res);
+
/* The complete format string was processed (with or without warnings). */
- return true;
+ return success;
}
/* Return the size of the object referenced by the expression DEST if
@@ -4184,9 +4664,18 @@ handle_printf_call (gimple_stmt_iterator *gsi, const vr_values *vr_values)
if (!info.fmtstr)
return false;
+ if (warn_restrict)
+ {
+ /* Compute the origin of the destination pointer and its offset
+ from the base object/pointer if possible. */
+ info.dst_offset = 0;
+ info.dst_origin = get_origin_and_offset (dstptr, &info.dst_field,
+ &info.dst_offset);
+ }
+
/* The result is the number of bytes output by the formatted function,
including the terminating NUL. */
- format_result res = format_result ();
+ format_result res;
/* I/O functions with no destination argument (i.e., all forms of fprintf
and printf) may fail under any conditions. Others (i.e., all forms of
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/builtin-sprintf-warn-23.c b/gcc/testsuite/gcc.dg/tree-ssa/builtin-sprintf-warn-23.c
new file mode 100644
index 00000000000..98dfa1d3966
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/builtin-sprintf-warn-23.c
@@ -0,0 +1,709 @@
+/* PR tree-optimization/35503 - Warning about restricted pointers?
+ { dg-do compile }
+ { dg-options "-O2 -Wno-format-overflow -Wrestrict -ftrack-macro-expansion=0" }
+*/
+
+void sink (int);
+
+#define S10 "0123456789"
+
+extern char a2[2][22];
+
+#define T(d, ...) do { \
+ char a[22] = S10; \
+ sink (__builtin_sprintf ((d), __VA_ARGS__)); \
+ } while (0)
+
+void test_ptr (char *d, int i)
+{
+ T (d, "%s", d); /* { dg-warning "argument 3 overlaps destination object 'd'" } */
+ T (d, "%s", d + 0); /* { dg-warning "overlaps" } */
+
+ /* The following only overlaps if d[1] is non-zero. */
+ T (d, "%s", d + 1); /* { dg-warning "may overlap" } */
+ T (d, "%s", d + 2); /* { dg-warning "may overlap" } */
+ T (d, "%s", d + i); /* { dg-warning "may overlap" } */
+
+ T (d, "%s", &d[0]); /* { dg-warning "overlaps" } */
+ T (d, "%s", &d[1]); /* { dg-warning "may overlap" } */
+ T (d, "%s", &d[2]); /* { dg-warning "may overlap" } */
+ T (d, "%s", &d[i]); /* { dg-warning "may overlap" } */
+
+ T (d + 0, "%s", d); /* { dg-warning "overlaps" } */
+ T (d + 1, "%s", d); /* { dg-warning "may overlap" } */
+ T (d + 2, "%s", d); /* { dg-warning "may overlap" } */
+ T (d + i, "%s", d); /* { dg-warning "may overlap" } */
+
+ T (&d[0], "%s", d); /* { dg-warning "overlaps" } */
+ T (&d[1], "%s", d); /* { dg-warning "may overlap" } */
+ T (&d[2], "%s", d); /* { dg-warning "may overlap" } */
+ T (&d[i], "%s", d); /* { dg-warning "may overlap" } */
+
+ const char *s = d;
+
+ T (d, "%s", s); /* { dg-warning "overlaps" } */
+ T (d, "%s", s + 1); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + 2); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + i); /* { dg-warning "may overlap" } */
+}
+
+void test_ptr_plus (char *d, int i)
+{
+ const char *s = d;
+
+ T (d, "%s", s++); /* { dg-warning "overlaps" } */
+ T (d, "%s", s++); /* { dg-warning "may overlap" } */
+ T (d, "%s", s++); /* { dg-warning "may overlap" } */
+ T (d, "%s", s++); /* { dg-warning "may overlap" } */
+
+ s += i;
+ T (d, "%s", s); /* { dg-warning "may overlap" } */
+}
+
+void test_array_1_dim (int i)
+{
+ T (a, "%s", a); /* { dg-warning "overlaps" } */
+ T (a, "%s", a + 0); /* { dg-warning "overlaps" } */
+ T (a, "%s", a + 1); /* { dg-warning "overlaps" } */
+ T (a, "%s", a + 2); /* { dg-warning "overlaps" } */
+ T (a, "%s", a + i); /* { dg-warning "may overlap" } */
+
+ T (a, "%s", &a[0]); /* { dg-warning "overlaps" } */
+ T (a, "%s", &a[1]); /* { dg-warning "overlaps" } */
+ T (a, "%s", &a[2]); /* { dg-warning "overlaps" } */
+ T (a, "%s", &a[i]); /* { dg-warning "may overlap" } */
+
+ T (a + 0, "%s", a); /* { dg-warning "overlaps" } */
+ T (a + 1, "%s", a); /* { dg-warning "overlaps" } */
+ T (a + 2, "%s", a); /* { dg-warning "overlaps" } */
+ T (a + i, "%s", a); /* { dg-warning "may overlap" } */
+
+ T (&a[0], "%s", a); /* { dg-warning "overlaps" } */
+ T (&a[1], "%s", a); /* { dg-warning "overlaps" } */
+ T (&a[2], "%s", a); /* { dg-warning "overlaps" } */
+ T (&a[i], "%s", a); /* { dg-warning "may overlap" } */
+}
+
+
+void test_array_2_dim (int i)
+{
+ T (a2[0], "%s", a2[0]); /* { dg-warning "overlaps" } */
+ T (a2[0], "%s", a2[0] + 0); /* { dg-warning "overlaps" } */
+ T (a2[0], "%s", a2[0] + 1); /* { dg-warning "may overlap" } */
+ T (a2[0], "%s", a2[0] + 2); /* { dg-warning "may overlap" } */
+ T (a2[0], "%s", a2[0] + i); /* { dg-warning "may overlap" } */
+
+ T (a2[0], "%s", &a2[0][0]); /* { dg-warning "overlaps" } */
+ T (a2[0], "%s", &a2[0][1]); /* { dg-warning "may overlap" } */
+ T (a2[0], "%s", &a2[0][2]); /* { dg-warning "may overlap" } */
+ T (a2[0], "%s", &a2[0][i]); /* { dg-warning "may overlap" } */
+
+ T (a2[0] + 0, "%s", a2[0]); /* { dg-warning "overlaps" } */
+ T (a2[0] + 1, "%s", a2[0]); /* { dg-warning "may overlap" } */
+ T (a2[0] + 2, "%s", a2[0]); /* { dg-warning "may overlap" } */
+ T (a2[0] + i, "%s", a2[0]); /* { dg-warning "may overlap" } */
+
+ T (&a2[0][0], "%s", a2[0]); /* { dg-warning "overlaps" } */
+ T (&a2[0][1], "%s", a2[0]); /* { dg-warning "may overlap" } */
+ T (&a2[0][2], "%s", a2[0]); /* { dg-warning "may overlap" } */
+ T (&a2[0][i], "%s", a2[0]); /* { dg-warning "may overlap" } */
+
+
+ T (a2[0], "%s", a2[1]);
+ T (a2[0], "%s", a2[1] + 0);
+ T (a2[0], "%s", a2[1] + 1);
+ T (a2[0], "%s", a2[1] + 2);
+ T (a2[0], "%s", a2[1] + i);
+
+ T (a2[0], "%s", &a2[1][0]);
+ T (a2[0], "%s", &a2[1][1]);
+ T (a2[0], "%s", &a2[1][2]);
+
+ /* a2[0] is represented as &a in Gimple, and &a2[1][i] as &a + _2,
+ with _1 defined to something like 10 + _1, and _1 to i. That
+ makes it virtually impossible to reliably determine that the
+ two pointers refer to distinct sub-arrays of the same multi-
+ dimensional array. */
+ T (a2[0], "%s", &a2[1][i]); /* { dg-bogus "overlap" "" { xfail *-*-* } } */
+
+ T (a2[0] + 0, "%s", a2[1]);
+ T (a2[0] + 1, "%s", a2[1]);
+ T (a2[0] + 2, "%s", a2[1]);
+ T (a2[0] + i, "%s", a2[1]);
+
+ T (&a2[0][0], "%s", a2[1]);
+ T (&a2[0][1], "%s", a2[1]);
+ T (&a2[0][2], "%s", a2[1]);
+ T (&a2[0][i], "%s", a2[1]);
+
+
+ T (a2[1], "%s", a2[0]);
+ T (a2[1], "%s", a2[0] + 0);
+ T (a2[1], "%s", a2[0] + 1);
+ T (a2[1], "%s", a2[0] + 2);
+ T (a2[1], "%s", a2[0] + i);
+
+ T (a2[1], "%s", &a2[0][0]);
+ T (a2[1], "%s", &a2[0][1]);
+ T (a2[1], "%s", &a2[0][2]);
+ T (a2[1], "%s", &a2[0][i]);
+
+ T (a2[1] + 0, "%s", a2[0]);
+ T (a2[1] + 1, "%s", a2[0]);
+ T (a2[1] + 2, "%s", a2[0]);
+ T (a2[1] + i, "%s", a2[0]);
+
+ T (&a2[1][0], "%s", a2[0]);
+ T (&a2[1][1], "%s", a2[0]);
+ T (&a2[1][2], "%s", a2[0]);
+ T (&a2[1][i], "%s", a2[0]); /* { dg-bogus "overlap" "" { xfail *-*-* } } */
+
+
+ T (a2[1], "%s", a2[1]); /* { dg-warning "overlaps" } */
+ T (a2[1], "%s", a2[1] + 0); /* { dg-warning "overlaps" } */
+ T (a2[1], "%s", a2[1] + 1); /* { dg-warning "may overlap" } */
+ T (a2[1], "%s", a2[1] + 2); /* { dg-warning "may overlap" } */
+ T (a2[1], "%s", a2[1] + i); /* { dg-warning "may overlap" } */
+
+ T (a2[1], "%s", &a2[1][0]); /* { dg-warning "overlaps" } */
+ T (a2[1], "%s", &a2[1][1]); /* { dg-warning "may overlap" } */
+ T (a2[1], "%s", &a2[1][2]); /* { dg-warning "may overlap" } */
+ T (a2[1], "%s", &a2[1][i]); /* { dg-warning "may overlap" "" { xfail *-*-* } } */
+
+ T (a2[1] + 0, "%s", a2[1]); /* { dg-warning "overlaps" } */
+ T (a2[1] + 1, "%s", a2[1]); /* { dg-warning "may overlap" } */
+ T (a2[1] + 2, "%s", a2[1]); /* { dg-warning "may overlap" } */
+ T (a2[1] + i, "%s", a2[1]); /* { dg-warning "may overlap" } */
+
+ T (&a2[1][0], "%s", a2[1]); /* { dg-warning "overlaps" } */
+ T (&a2[1][1], "%s", a2[1]); /* { dg-warning "may overlap" } */
+ T (&a2[1][2], "%s", a2[1]); /* { dg-warning "may overlap" } */
+ T (&a2[1][i], "%s", a2[1]); /* { dg-warning "may overlap" "" { xfail *-*-* } } */
+}
+
+struct S {
+ char a[4];
+ char b[4];
+};
+
+struct S2 {
+ struct S s_1;
+ struct S s_2;
+ struct S sa3[3];
+};
+
+struct S3 {
+ struct S2 s2_1;
+ struct S2 s2_2;
+
+ struct {
+ struct {
+ struct {
+ struct S sa_3[3];
+ } a_1[3];
+ } a_2[3][3];
+ } a_3[3][3][3];
+
+ char fa[];
+};
+
+void test_struct_member_array (struct S3 *s3, int i)
+{
+ char *d = s3->s2_1.s_1.a;
+
+ T (d, "%s", d); /* { dg-warning "overlaps" } */
+ T (d, "%s", d + 0); /* { dg-warning "overlaps" } */
+ T (d, "%s", d + 1); /* { dg-warning "may overlap" } */
+ T (d, "%s", d + 2); /* { dg-warning "may overlap" } */
+ T (d, "%s", d + i); /* { dg-warning "may overlap" } */
+
+ T (d, "%s", &d[0]); /* { dg-warning "overlaps" } */
+ T (d, "%s", &d[1]); /* { dg-warning "may overlap" } */
+ T (d, "%s", &d[2]); /* { dg-warning "may overlap" } */
+ T (d, "%s", &d[i]); /* { dg-warning "may overlap" } */
+
+ T (d + 0, "%s", d); /* { dg-warning "overlaps" } */
+ T (d + 1, "%s", d); /* { dg-warning "may overlap" } */
+ T (d + 2, "%s", d); /* { dg-warning "may overlap" } */
+ T (d + i, "%s", d); /* { dg-warning "may overlap" } */
+
+ T (&d[0], "%s", d); /* { dg-warning "overlaps" } */
+ T (&d[1], "%s", d); /* { dg-warning "may overlap" } */
+ T (&d[2], "%s", d); /* { dg-warning "may overlap" } */
+ T (&d[i], "%s", d); /* { dg-warning "may overlap" } */
+
+ const char *s = d;
+
+ T (d, "%s", s); /* { dg-warning "overlaps" } */
+ T (d, "%s", s + 1); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + 2); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + i); /* { dg-warning "may overlap" } */
+
+ s = s3->s2_1.s_1.b;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = s3->s2_1.s_2.a;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = s3->s2_1.s_2.b;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ d = s3->s2_1.s_1.b;
+
+ s = s3->s2_1.s_2.a;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = s3->s2_1.s_2.b;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = s3->s2_2.s_1.a;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = s3->s2_2.s_1.b;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = s3->s2_2.s_2.a;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = s3->s2_2.s_2.b;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ d = s3->s2_2.s_1.a;
+
+ s = s3->s2_1.s_1.a;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = s3->s2_1.s_1.b;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = s3->s2_2.s_1.a;
+
+ T (d, "%s", s); /* { dg-warning "overlaps" } */
+ T (d, "%s", s + 1); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + 2); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + i); /* { dg-warning "may overlap" } */
+
+ s = s3->s2_2.s_2.a;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = s3->s2_2.s_2.b;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ d = s3->s2_2.s_1.b;
+
+ s = s3->s2_1.s_1.a;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = s3->s2_1.s_1.b;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = s3->s2_2.s_1.a;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = s3->s2_2.s_1.b;
+
+ T (d, "%s", s); /* { dg-warning "overlaps" } */
+ T (d, "%s", s + 1); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + 2); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + i); /* { dg-warning "may overlap" } */
+
+ s = s3->s2_2.s_2.a;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = s3->s2_2.s_2.b;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+}
+
+void test_struct_member_array_array (struct S3 *s3, int i)
+{
+ char *d = s3->s2_1.sa3[0].a;
+ char *s = s3->s2_1.sa3[0].a;
+
+ T (d, "%s", s); /* { dg-warning "overlaps" } */
+ T (d, "%s", s + 0); /* { dg-warning "overlaps" } */
+ T (d, "%s", s + 1); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + 2); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + i); /* { dg-warning "may overlap" } */
+
+ T (d, "%s", &s[0]); /* { dg-warning "overlaps" } */
+ T (d, "%s", &s[1]); /* { dg-warning "may overlap" } */
+ T (d, "%s", &s[2]); /* { dg-warning "may overlap" } */
+ T (d, "%s", &s[i]); /* { dg-warning "may overlap" } */
+
+ T (d + 0, "%s", s); /* { dg-warning "overlaps" } */
+ T (d + 1, "%s", s); /* { dg-warning "may overlap" } */
+ T (d + 2, "%s", s); /* { dg-warning "may overlap" } */
+ T (d + i, "%s", s); /* { dg-warning "may overlap" } */
+
+ T (&d[0], "%s", s); /* { dg-warning "overlaps" } */
+ T (&d[1], "%s", s); /* { dg-warning "may overlap" } */
+ T (&d[2], "%s", s); /* { dg-warning "may overlap" } */
+ T (&d[i], "%s", s); /* { dg-warning "may overlap" } */
+
+ s = s3->s2_1.sa3[0].b;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 0);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ T (d, "%s", &s[0]);
+ T (d, "%s", &s[1]);
+ T (d, "%s", &s[2]);
+ T (d, "%s", &s[i]);
+
+ T (d + 0, "%s", s);
+ T (d + 1, "%s", s);
+ T (d + 2, "%s", s);
+ T (d + i, "%s", s);
+
+ T (&d[0], "%s", s);
+ T (&d[1], "%s", s);
+ T (&d[2], "%s", s);
+ T (&d[i], "%s", s);
+}
+
+void test_struct_member_nested_array (struct S3 *s3, int i)
+{
+#define PFX(sfx) s3->a_3[i3_0][i3_1][i3_2] \
+ .a_2[i2_0][i2_1] \
+ .a_1[i1_0].sfx
+
+#define TT(d, s) \
+ T (PFX (d), "%s", PFX (s)); \
+ T (PFX (d), "%s", &PFX (s)[0]); \
+ T (PFX (d), "%s", &PFX (s)[1]); \
+ T (PFX (d), "%s", &PFX (s)[2]); \
+ T (PFX (d), "%s", &PFX (s)[i]) \
+
+#define T1(i0) do \
+ { \
+ enum {i1_0 = i0 }; \
+ \
+ TT (sa_3[0].a, sa_3[0].b); TT (sa_3[0].b, sa_3[0].a); \
+ \
+ TT (sa_3[0].a, sa_3[1].a); TT (sa_3[0].a, sa_3[1].b); \
+ TT (sa_3[0].b, sa_3[1].a); TT (sa_3[0].b, sa_3[1].b); \
+ \
+ TT (sa_3[0].a, sa_3[2].a); TT (sa_3[0].a, sa_3[2].b); \
+ TT (sa_3[0].b, sa_3[2].a); TT (sa_3[0].b, sa_3[2].b); \
+ \
+ TT (sa_3[1].a, sa_3[0].a); TT (sa_3[1].a, sa_3[0].b); \
+ TT (sa_3[1].b, sa_3[0].a); TT (sa_3[1].b, sa_3[0].b); \
+ \
+ TT (sa_3[1].a, sa_3[1].b); TT (sa_3[1].b, sa_3[1].a); \
+ \
+ TT (sa_3[1].a, sa_3[2].a); TT (sa_3[1].a, sa_3[2].b); \
+ TT (sa_3[1].b, sa_3[2].a); TT (sa_3[1].b, sa_3[2].b); \
+ } while (0)
+
+#define T2(i0, i1) do \
+ { \
+ enum { i2_0 = i0, i2_1 = i1 }; \
+ T1 (0); T1 (1); T1 (2); \
+ } while (0)
+
+#define T3(i0, i1, i2) do \
+ { \
+ enum { i3_0 = i0, i3_1 = i1, i3_2 = i2 }; \
+ T2 (0, 0); T2 (0, 1); T2 (0, 2); \
+ T2 (1, 0); T2 (1, 1); T2 (1, 2); \
+ T2 (2, 0); T2 (2, 1); T2 (2, 2); \
+ } while (0)
+
+#if 0
+ /* These tests take forever and a day to compile. Enable them only
+ during the development of this feature but leave them otherwise
+ disabled to avoid slowing everything down for others. */
+ T3 (0, 0, 0); T3 (0, 0, 1); T3 (0, 0, 2);
+ T3 (0, 1, 0); T3 (0, 1, 1); T3 (0, 1, 2);
+ T3 (0, 2, 0); T3 (0, 2, 1); T3 (0, 2, 2);
+
+ T3 (1, 0, 0); T3 (1, 0, 1); T3 (1, 0, 2);
+ T3 (1, 1, 0); T3 (1, 1, 1); T3 (1, 1, 2);
+ T3 (1, 2, 0); T3 (1, 2, 1); T3 (1, 2, 2);
+
+ T3 (2, 0, 0); T3 (2, 0, 1); T3 (2, 0, 2);
+ T3 (2, 1, 0); T3 (2, 1, 1); T3 (2, 1, 2);
+ T3 (2, 2, 0); T3 (2, 2, 1); T3 (2, 2, 2);
+#endif
+}
+
+void test_struct_member_flexarray (struct S3 *s3, int i, int j)
+{
+ char *d = s3->fa;
+ char *s = s3->fa;
+
+ T (d, "%s", s); /* { dg-warning "overlaps" } */
+ T (d, "%s", s + 0); /* { dg-warning "overlaps" } */
+ T (d, "%s", s + 1); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + 2); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + i); /* { dg-warning "may overlap" } */
+
+ T (d, "%s", &s[0]); /* { dg-warning "overlaps" } */
+ T (d, "%s", &s[1]); /* { dg-warning "may overlap" } */
+ T (d, "%s", &s[2]); /* { dg-warning "may overlap" } */
+ T (d, "%s", &s[i]); /* { dg-warning "may overlap" } */
+
+ T (d + 0, "%s", s); /* { dg-warning "overlaps" } */
+ T (d + 1, "%s", s); /* { dg-warning "may overlap" } */
+ T (d + 2, "%s", s); /* { dg-warning "may overlap" } */
+ T (d + i, "%s", s); /* { dg-warning "may overlap" } */
+
+ T (&d[0], "%s", s); /* { dg-warning "overlaps" } */
+ T (&d[1], "%s", s); /* { dg-warning "may overlap" } */
+ T (&d[2], "%s", s); /* { dg-warning "may overlap" } */
+ T (&d[i], "%s", s); /* { dg-warning "may overlap" } */
+
+ d = s3->fa + i;
+ s = s3->fa + j;
+
+ T (d, "%s", s); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + 0); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + 1); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + 2); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + i); /* { dg-warning "may overlap" } */
+
+ T (d, "%s", &s[0]); /* { dg-warning "may overlap" } */
+ T (d, "%s", &s[1]); /* { dg-warning "may overlap" } */
+ T (d, "%s", &s[2]); /* { dg-warning "may overlap" } */
+ T (d, "%s", &s[i]); /* { dg-warning "may overlap" } */
+
+ T (d + 0, "%s", s); /* { dg-warning "may overlap" } */
+ T (d + 1, "%s", s); /* { dg-warning "may overlap" } */
+ T (d + 2, "%s", s); /* { dg-warning "may overlap" } */
+ T (d + j, "%s", s); /* { dg-warning "may overlap" } */
+
+ T (&d[0], "%s", s); /* { dg-warning "may overlap" } */
+ T (&d[1], "%s", s); /* { dg-warning "may overlap" } */
+ T (&d[2], "%s", s); /* { dg-warning "may overlap" } */
+ T (&d[j], "%s", s); /* { dg-warning "may overlap" } */
+}
+
+union U
+{
+ struct S2 s2_1;
+ struct S2 s2_2;
+};
+
+void test_union_member_array (union U *un, int i)
+{
+ char *d = un->s2_1.s_1.a;
+
+ T (d, "%s", d); /* { dg-warning "overlaps" } */
+ T (d, "%s", d + 0); /* { dg-warning "overlaps" } */
+ T (d, "%s", d + 1); /* { dg-warning "may overlap" } */
+ T (d, "%s", d + 2); /* { dg-warning "may overlap" } */
+ T (d, "%s", d + i); /* { dg-warning "may overlap" } */
+
+ T (d, "%s", &d[0]); /* { dg-warning "overlaps" } */
+ T (d, "%s", &d[1]); /* { dg-warning "may overlap" } */
+ T (d, "%s", &d[2]); /* { dg-warning "may overlap" } */
+ T (d, "%s", &d[i]); /* { dg-warning "may overlap" } */
+
+ T (d + 0, "%s", d); /* { dg-warning "overlaps" } */
+ T (d + 1, "%s", d); /* { dg-warning "may overlap" } */
+ T (d + 2, "%s", d); /* { dg-warning "may overlap" } */
+ T (d + i, "%s", d); /* { dg-warning "may overlap" } */
+
+ T (&d[0], "%s", d); /* { dg-warning "overlaps" } */
+ T (&d[1], "%s", d); /* { dg-warning "may overlap" } */
+ T (&d[2], "%s", d); /* { dg-warning "may overlap" } */
+ T (&d[i], "%s", d); /* { dg-warning "may overlap" } */
+
+ const char *s = d;
+
+ T (d, "%s", s); /* { dg-warning "overlaps" } */
+ T (d, "%s", s + 1); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + 2); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + i); /* { dg-warning "may overlap" } */
+
+ s = un->s2_1.s_1.b;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = un->s2_1.s_2.a;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = un->s2_1.s_2.b;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+
+ s = un->s2_2.s_1.a;
+
+ T (d, "%s", s); /* { dg-warning "overlaps" } */
+ T (d, "%s", s + 1); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + 2); /* { dg-warning "may overlap" } */
+ T (d, "%s", s + i); /* { dg-warning "may overlap" } */
+
+ s = un->s2_2.s_1.b;
+
+ T (d, "%s", s);
+ T (d, "%s", s + 1);
+ T (d, "%s", s + 2);
+ T (d, "%s", s + i);
+}
+
+void test_multiple_overlap (int i)
+{
+ {
+ char a[3] = ""; /* { dg-message "declared here" } */
+
+ /* Both a3 and a4 definitely overlap the output even if the are
+ empty because of the terminating nul. */
+ char *d = a;
+ char *a3 = a + 0;
+ char *a4 = a - 0;
+
+ T (d, "%s%s", a3, a4); /* { dg-warning "arguments 3, 4 overlap destination object .a." }*/
+ }
+
+ {
+ char a[4]; /* { dg-message "declared here" } */
+
+ /* a3 and a4 may overlap the output. They will only not overlap
+ it when a3 is empty, and a4 is at most chaeracter byte long. */
+ char *d = a;
+ char *a3 = a + 2;
+ char *a4 = a + 3;
+
+ T (d, "%s%s", a3, a4); /* { dg-warning "arguments 3, 4 may overlap destination object .a." }*/
+ }
+
+ {
+ char a[5]; /* { dg-message "declared here" } */
+
+ /* a3 certaibly overlaps the output, but a4 may or may not depending
+ in the value of i. */
+ char *d = a;
+ char *a3 = a + 0;
+ char *a4 = a + i;
+
+ T (d, "%s%s", a3, a4); /* { dg-warning "arguments 3 and maybe 4 overlap destination object .a." }*/
+ }
+}
+
+void test_overlap_with_precision (char *d, int i, int j)
+{
+ /* None of the following is diagnosed because no copying takes place
+ between the %s argument and the destination. */
+ T (d, "%.0s", d + 0);
+ T (d, "%.0s", d + 1);
+ T (d, "%.0s", d + 2);
+ T (d, "%.0s", d + i);
+
+ T (d, "%.1s", d + 0); /* { dg-warning "overlaps" } */
+ /* Unlike the %.0s case the following deserves a warning because
+ when d[1] isn't nul, it will be overwritten by the terminating
+ nul added by sprintf. */
+ T (d, "%.1s", d + 1); /* { dg-warning "may overlap" } */
+ T (d, "%.1s", d + 2);
+ T (d, "%.1s", d + i); /* { dg-warning "may overlap" } */
+
+ T (d + 1, "%.0s", d);
+ T (d + 2, "%.0s", d);
+
+ T (d + 1, "%.1s", d); /* { dg-warning "may overlap" } */
+ T (d + 2, "%.1s", d);
+
+ T (d + 2, "%.1s", d + 1); /* { dg-warning "may overlap" } */
+ T (d + 2, "%.1s", d + i); /* { dg-warning "may overlap" } */
+
+ /* The following should be "overlaps" but tracking that the offset
+ is the same variable doesn't seem worth the effort. */
+ T (d + i, "%.1s", d + i); /* { dg-warning "overlap" } */
+
+ T (d + i, "%.1s", d + j); /* { dg-warning "may overlap" } */
+
+ /* Exercise overlap with precison in the range [0, 1]. */
+ if (i < 0 || 1 < i)
+ i = 0;
+
+ T (d, "%.*s", i, d + 0); /* { dg-warning "may overlap" } */
+ T (d, "%.*s", i, d + 1); /* { dg-warning "may overlap" } */
+ T (d, "%.*s", i, d + 2);
+ T (d, "%.*s", i, d + i); /* { dg-warning "may overlap" "" { xfail *-*-* } } */
+}
diff --git a/gcc/tree-ssa-strlen.c b/gcc/tree-ssa-strlen.c
index c82bc7c4815..8e2f98d31f2 100644
--- a/gcc/tree-ssa-strlen.c
+++ b/gcc/tree-ssa-strlen.c
@@ -5016,6 +5016,7 @@ pass_strlen::gate (function *)
{
return ((warn_format_overflow > 0
|| warn_format_trunc > 0
+ || warn_restrict > 0
|| flag_optimize_strlen > 0
|| flag_printf_return_value)
&& optimize > 0);