tst-posix_fallocate test failure

Ryan Arnold ryan.arnold@gmail.com
Wed Nov 26 19:36:00 GMT 2008


On Wed, Nov 26, 2008 at 12:52 PM, Renaud Marquet <rmarquet@gmail.com> wrote:
> Le mercredi 26 novembre 2008 à 09:41 -0500, Carlos O'Donell a écrit :
>> On Tue, Nov 25, 2008 at 5:23 PM, Renaud Marquet <rmarquet@gmail.com> wrote:
>> > As suggested I did a little step into the failing test. But I cannot
>> > figure out what's happening. I hope you'll get a clearer idea of the
>> > problem with this. Here's a dump of the gdb session :
>>
>> Nothing seems out of place, it doesn't look like the syscall to fallocate fails.
>>
>> You'll have to figure out what the test is trying to assert and see if
>> it is logically true.
>
> When I link the test with host system's glibc, it do not fail although
> it's also a PowerPC 64 bits binary. So I can reasonably think the test
> is correct and should complete successfully. It's true syscall don't
> fail (no error returned) but at the same time it seems to do absolutely
> nothing (file size don't change). As as said I don't understand why the
> code in lines 42 and 44 seems to be executed twice. Is it related to
> some macro expansion ?

You see it twice because GDB is an assembler level debugger and
there's not a direct mapping between C code and assembler instruction
progression.  Therefore, if a line of C code is implemented in several
assembler operations you may see it come up twice.  It'll come up out
of order as well if the compiler has generated assembler intermixed
with assembler from other lines of C-code for code dispatch
optimization.

>> Perhaps one of the PowerPC users has seen this test failure and knows
>> the reason.
>>
>> Have you looked through the bugzilla? Filed a bugzilla bug?
>
> I already looked over google and bugzilla before posting here and did
> not find any other report of this problem. I don't think I'll file a new
> bug as it seems to be very specific to my build and I didn't test it
> against latest cvs version.
>
> This was the only test that failed and it's not really a critical
> function. Moreover calling the function produces no error. So I guess
> I'll try to live with it hoping it will not become a serious issue
> later.
>
> Anyway thank you for your help and suggestions.

I'm the defacto PowerPC maintainer at this point.

I'll check-out a tree and configure with the options you provide.

Ryan S. Arnold


More information about the Libc-help mailing list