This is the mail archive of the
mailing list for the GCC project.
Re: Anyone got irix5? I have questions about your ctype implementation
> From: Rainer Orth <ro@TechFak.Uni-Bielefeld.DE>
> Kaveh R. Ghazi writes:
> > Listed in three categories, based on whether I have a solution and if
> > it has been installed, the problems I've worked on or am working on are:
> Thanks for this overview: it proved very useful and allowed me to make some
> progress on the unresolved failures below.
> > 7. libstdc++-v3 failures on irix6.2 beyond those that appear on 6.5.
> > FAIL: 21_strings/ctor_copy_dtor.cc execution test
> > FAIL: 21_strings/insert.cc execution test
> These two are due to a bug in the testsuite: setrlimit was called in a way
> which caused the hard limits to be reduced below the soft limits, thus all
> calls to malloc fail ;-( I'll post a patch shortly.
Excellent detective work! I had tracked it down to malloc returning
NULL stepping through using gdb but couldn't figure out why it failed.
> > FAIL: 27_io/ios_base_storage.cc execution test
> This one is really absurd: malloc happens to be called with
> (size_t)(8*(INT_MAX-1)) (4294967288)
> but instead of returning NULL on a 32-bit machine with way below 4 GB swap,
> it returns a valid pointer, but later accesses to that memory area cause
> the testcase to crash. Linking with -lmalloc instead of using the version
> of malloc in libc allows this test to succeed. I fear we can do nothing
> about this if SGI is unable to implement malloc correctly ;-(
Ok, thanks for checking. Perhaps this one should get xfailed for irix
less than version 6.5? (I'm not sure when the malloc bug was fixed.
Anyone got 6.3 or 6.4 who can test this?)
Kaveh R. Ghazi Director of Systems Architecture
email@example.com Qwest Global Services