Re: [RFA:] ar test failure fix

[Mark Mitchell <mark at codesourcery dot com>]

Hans-Peter Nilsson wrote:
> > Date: Tue, 07 Jun 2005 19:51:16 -0700
> > From: Mark Mitchell <mark@codesourcery.com>
>
>
> > Hans-Peter Nilsson wrote:
> >
> > > Mark, did you really run the testsuite after your bfd-fopen ABI
> > > changes, in a *clean* directory (no partial test-results or
> > > left-over files)?  My autotester for cris-elf did, and got
> > > (after the recent new-port-commit debacle and switch braino was
> > > fixed, that is):
> >
> > I ran the GDB testsuite, and built Binutils with --enable-target=all. 
> > Was I supposed to run the binutils testsuite too?
>
>
> When you change anything in binutils, including bfd, yes.

OK.

> My patch is for bfd_fopen which only did that on error, not bfd_fdopenr.
>
> So there *is* a latent bug.

I'm not sure what you mean by "*is*".  Do you mean that yes, bfd_fdopenr 
was always buggy, and now I've broken more stuff?  Or that bfd_fdopenr 
was OK, but now I've broken more stuff?  In any case, your patch seems 
entirely reasonable to me, but I can't approve it.  And, I'm not sure if 
it's obvious given that the code seems to have been broken before. 
Perhaps, even, the caller is buggy?  The documentation for these 
functions does say that "if NULL is returned then an error has 
occurred", which could be read as saying that if non-NULL is returned, 
then there is no error, so you shouldn't be checking for on?

In the short term, should I revert my patch until this is sorted?

Thanks,

--
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com
(916) 791-8304