[Bug 71043] Use STACK lint tool to clean code ...

https://bugs.freedesktop.org/show_bug.cgi?id=71043

Stephan Bergmann <sbergman@redhat.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |sbergman@redhat.com

--- Comment #12 from Stephan Bergmann <sbergman@redhat.com> ---
(In reply to comment #5)
> Now, what to do in this particular case then is a question of taste, more or
> less.
>
> Personally I think that if a memory allocation (of a small amount of memory)
> fails at this place in the code, which as far as I know is invoked mainly
> (only?) very early when LO is starting, something is very wrong and there is
> no point in even trying to recover gracefully from the memory allocation
> failure. So let __osl_createPipeImpl() be as is and remove the pointless
> checks for NULL return after the calls to it.
>
> But I assume there is an opposite opinion, too, that each and every memory
> allocation should be checked and the code should do its utmost to fail
> gracefully... In that case, the check for memory allocation failure should
> be moved inside __osl_createPipeImpl() right after the call to calloc(), and
> in case of failure, NULL should be returned immediately. Hopefully then the
> return value of __osl_createPipeImpl() is checked in each case.

In this case, where both callers of __osl_createPipeImpl already use "return
NULL" to indicate error, I see no disadvantage in cleanly reporting calloc
failure instead of relying on SIGSEGV doing the right thing.

-- 
You are receiving this mail because:
You are on the CC list for the bug.