This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc 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: Corrupted test cases since "Don't require test wrappers to preserve environment variables, use more consistent environment" heaps


On Tue, 24 Feb 2015, Mel Gorman wrote:

> The test case does not crash but it's clearly corrupted.
> 
> Before that commit we have
> 
> mel@stampy:~/git-public/glibc > rm -rf obj-verify; git checkout 8540f6d2a74fe9d67440535ebbcfa252180a3172^ | head; mkdir obj-verify; cd obj-verify; ../configure --prefix=/tmp/blah > /dev/null 2>&1 && make -j8 > /dev/null 2>&1 && make -j8 tests > /dev/null 2>&1 ; ./testrun.sh > ./intl/tst-gettext4; cd ..
> HEAD is now at ed36bfa18faf... PowerPC: Fix optimized strncat strlen call
> beauty
> thread 1 call 1 returned: beauty
> beauty
> thread 2 call 1 returned: beauty
> beauty
> thread 1 call 2 returned: beauty
> beauty
> thread 2 call 2 returned: beauty

This is the output when the test fails; "thread 1 call 1 returned" etc. 
are failure messages.  Before my commit, testrun.sh was not by itself a 
correct way of running this test, because it lacked the environment setup 
required by this test; that environment setup was local to the 
tst-gettext4.sh script.

> And on the commit we get
> mel@stampy:~/git-public/glibc > rm -rf obj-verify; git checkout 8540f6d2a74fe9d67440535ebbcfa252180a3172 | head; mkdir obj-verify; cd obj-verify; ../configure --prefix=/tmp/blah > /dev/null 2>&1 && make -j8 > /dev/null 2>&1 && make -j8 tests > /dev/null 2>&1 ; ./testrun.sh > ./intl/tst-gettext4; cd ..
> Previous HEAD position was ed36bfa18faf... PowerPC: Fix optimized strncat strlen call
> HEAD is now at 8540f6d2a74f... Don't require test wrappers to preserve environment variables, use more consistent environment.
> Schï¿œnheit
> beautᅵ
> Schï¿œnheit
> beautᅵ

This is passing output (modulo a few character set conversions after the 
test produced its output in ISO-8859-1, but the test clearly worked 
because it didn't produce any of the "returned" messages from failed 
comparisons).  Because the environment setup is now shared for all tests 
in the makefiles, after my commit testrun.sh is suitable to use to run 
this test, and so it runs with the correct passing output.

-- 
Joseph S. Myers
joseph@codesourcery.com

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