This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] asan unit tests from llvm lit-test


On Wed, Nov 28, 2012 at 02:40:55PM +0400, Konstantin Serebryany wrote:
> I fully agree about "minimal testsuite".
> But, for example, porting the asan's gtest test (2+ KLOC) to another
> harness is probably too much.

Depends on how significant changes to the test body are actually needed,
and if we could e.g. write a script that transforms the gtest test into
dejagnu test.
Say something minimal, like for each
  const char *uaf_string = "AddressSanitizer:.*heap-use-after-free";
  EXPECT_DEATH(uaf_test<U1>(1, 0), uaf_string);
replace that with
  DIE_IF(77, uaf_test<U1>(1, 0)); /* { dg-final { asan-die-if 77 "AddressSanitizer:.*heap-use-after-free" } } */
which would be essnetially
int die_if;
and at the beginning of main
  char *p = getenv ("ASAN_DIE_IF");
  if (p)
    die_if = atoi (p);
or so, then
#define DIE_IF(id, what) if (die_if == id) { what; }
where the test would be run normally first, then asan-die-if would
run it again (see e.g. gdb-test in guality.exp how it invokes gdb on the
test) with setenv ASAN_DIE_IF 77 (environment only to cope with target
boards that don't pass arguments, perhaps we could just ignore bare metal
targets for these kind of tests), and scan the output for the given regexp.
Problem with that is that unfortunately the regexps are runtime constructed,
aren't present as literals.
Or even slighly more involved solution would be to define
#define EXPECT_DEATH(x, y) \
  if (die_if == 0) \
    { \
      fprintf (stderr, "EXPECT_DEATH%d %s EXPECT_DEATHEND%d\n", \
	       die_if_counter, y, die_if_counter++); \
    } \
  else if (die_if_counter++ == die_if) \
    x
Then the test would be run once without ASAN_DIE_IF in environment (or =0),
that would produce output full of
EXPECT_DEATH1 AddressSanitizer:.*heap-use-after-free EXPECT_DEATHEND1
...
which tcl could parse, and figure from it that it should run the test
again 156 or how many times, with ASAN_DIE_IF from 1 to 156, and at each
iteration try to match the output against the regexp for that iteration.
Then you'd essentially just have to tweak a few lines at the start of the
test, includes, first few lines in main and that would be it.

	Jakub


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]