"make check--target-libstdc++-v3" failing for cross-compiler
Dan Kegel
dank@kegel.com
Sun Jul 13 19:56:00 GMT 2003
Phil Edwards wrote:
> On Sun, Jul 13, 2003 at 10:53:28AM -0700, Dan Kegel wrote:
>
>>That's correct from the point of view of the library code being
>>tested, I guess, but since I only have an x86 -> ppc cross
>>compiler, the compiler has to run on x86, and there the compiler
>>is definitely named ${TARGET}-g++, not just plain g++.
>
>
> Did you set RUNTESTFLAGS?
>
> http://gcc.gnu.org/simtest-howto.html
> http://gcc.gnu.org/install/test.html
Yes. My commandline was
RUNTESTFLAGS="--target=$TARGET -v -v -v -v -a" make check-target-libstdc++-v3
The following patch seems to get things working. This is related
to the patch I posted yesterday for check-g++ on gcc-patches.
Comments welcome.
------ gcc-3.3-libstdc++-v3-dg.exp.patch ---------
Without this patch, the command to test a cross-compiler's libstdc++-v3,
RUNTESTFLAGS=--target=powerpc-750-linux-gnu -v -v -v -v make check-target-libstdc++-v3
fails in two ways:
1. the -L option meant to locate the testsuite directory is incorrect, and
2. the wrong compiler is invoked, causing all sorts of havoc, not least of which
is the native compiler is invoked when we really wanted to invoke the cross-compiler
we just built.
Here's an example log of the problem in action. Every testcase fails, this shows just one:
-------------
Invoking the compiler as g++ -ggdb3 -DDEBUG_ASSERT -I/home3/dank/wk/ixos/dap_interim/linux/3rdParty/crosstool/crosstool-0.12/build/powerpc-750-linux-gnu/gcc-3.3-glibc-2.2.5/gcc-3.3/libstdc++-v3/testsuite
/home3/dank/wk/ixos/dap_interim/linux/3rdParty/crosstool/crosstool-0.12/build/powerpc-750-linux-gnu/gcc-3.3-glibc-2.2.5/gcc-3.3/libstdc++-v3/testsuite/20_util/binders.cc -g -O2 -DDEBUG_ASSERT -L/testsuite -lv3test -lm -o ./binders.exe
compiler exited with status 1
output is:
/usr/lib/gcc-lib/i386-linux/2.95.4/../../../../include/g++-3/stl_function.h: In method `void binder2nd<mem_fun1_ref_t<void,Elem,int> >::operator ()(const Elem &) const':^M
/usr/lib/gcc-lib/i386-linux/2.95.4/../../../../include/g++-3/stl_algo.h:83: instantiated from `for_each<Elem *, binder2nd<mem_fun1_ref_t<void,Elem,int> > >(Elem *, Elem *, binder2nd<mem_fun1_ref_t<void,Elem,int> >)'^M
/home3/dank/wk/ixos/dap_interim/linux/3rdParty/crosstool/crosstool-0.12/build/powerpc-750-linux-gnu/gcc-3.3-glibc-2.2.5/gcc-3.3/libstdc++-v3/testsuite/20_util/binders.cc:43: instantiated from here^M
/usr/lib/gcc-lib/i386-linux/2.95.4/../../../../include/g++-3/stl_function.h:224: conversion from `const Elem' to `Elem &' discards qualifiers^M
/usr/lib/gcc-lib/i386-linux/2.95.4/../../../../include/g++-3/stl_function.h:622: in passing argument 1 of `mem_fun1_ref_t<void,Elem,int>::operator ()(Elem &, int) const'^M
...
FAIL: 20_util/binders.cc (test for excess errors)
WARNING: 20_util/binders.cc compilation failed to produce executable
-------------
And here's the patch. I'm not happy with it, and it probably gets some cases wrong,
but it seems to work for the common native case and for my cross-compiler case.
--- gcc-3.3/libstdc++-v3/testsuite/lib/libstdc++-v3-dg.exp.old Sun Jul 13 10:42:01 2003
+++ gcc-3.3/libstdc++-v3/testsuite/lib/libstdc++-v3-dg.exp Sun Jul 13 11:39:54 2003
@@ -46,8 +46,23 @@
global gluefile wrap_flags
global ld_library_path
global tool_root_dir
+ global target_triplet
set blddir [lookfor_file [get_multilibs] libstdc++-v3]
+ if { $blddir == "" } {
+ set multilibs [get_multilibs]
+ # FIXME: assume multilib only one level deep
+ set multisub [file tail $multilibs]
+ verbose "libstdc++-v3-init: couldn't find libstdc++-v3 in $multilibs, trying $objdir"
+ set blddir [lookfor_file ${objdir} "$target_triplet/$multisub/libstdc++-v3"]
+ }
+ if { $blddir == "" } {
+ verbose "libstdc++-v3-init: couldn't find libstdc++-v3, trying $objdir without multilibs"
+ set blddir [lookfor_file ${objdir} "$target_triplet/libstdc++-v3"]
+ }
+ if { $blddir == "" } {
+ error "Can't find libstdc++-v3"
+ }
# By default, we assume we want to run program images.
global dg-do-what-default
--
Dan Kegel
http://www.kegel.com
http://counter.li.org/cgi-bin/runscript/display-person.cgi?user=78045
More information about the Gcc
mailing list