RE: [RFA] Fix display of array of unspecified length inside structures

["Pierre Muller" <pierre.muller@ics-cnrs.unistra.fr>]

> -----Message d'origine-----
> De : gdb-patches-owner@sourceware.org [mailto:gdb-patches-
> owner@sourceware.org] De la part de Pedro Alves
> Envoyé : vendredi 18 février 2011 16:25
> À : Pierre Muller
> Cc : gdb-patches@sourceware.org
> Objet : Re: [RFA] Fix display of array of unspecified length inside
> structures
> 
> On Friday 18 February 2011 14:54:03, Pierre Muller wrote:
> 
> > > On Friday 18 February 2011 11:08:38, Pierre Muller wrote:
> > >
> > > > PS: It could be wise to add some test in the testsuite for
> > > > this, but I have no idea where I could insert this kind of test,
> > > > any ideas?
> > >
> > > Yes, please.  We have surprisingly few tests for this sort of
> > > thing, AFAICS.  I'm not even sure this is a regression from
> > > my recent changes, I think it may well not be.
> >
> >   I checked out gdb version 7.2 shows this regression,
> > as compared to Cygwin 6.8 at least...
> > Which means that the regression is not really recent.
> >
> >   This might means that we should also merge this patch to
> > the branch, no?
> 
> Sound fine to me.

  Tested, the problem existed on 7.2 branch
and was fixed by the patch, thus I applied
the patch to 7.2 branch.

Pierre

> > > Zero-length arrays (as poor man's flexible arrays) are supported
> > > in GNU C as an extension.  To be portable, you'd
> > > need to use an array of length 1 (or c99's real flexible arrays),
> > > but that won't trigger the bug.
> >   Apparently there is also the flexible array member construct
> > see
> > http://gcc.gnu.org/onlinedocs/gcc-4.1.2/gcc/Zero-Length.html
> 
> That just confirms what I said.  :-)  The flexible array
> member construst is C99 only, so it's likely that other
> compilers choke on it by default.

  But this means that other compilers
will at least accept this flexible array construct if
they conform to ISO C99.

 
> > > I'd point at printcmds.exp, but I'm not sure if there are compilers
> > > out there that choke on the construct...  There's always a
> > > new test file option...
> > >
> > > > PS2: It is probably impossible to make such a test without
> > > > alloca or some other memory allocation function, no?
> > > > Are there any system restriction for this?
> >
> >   There is a long check at start of gdb.base/funcargs.c
> > but it might just be to really check that alloca really uses
> > the stack...
> 
> Irk.  Just use malloc then?  It's not really crutial that
> the test runs on all targets/hosts.  As long as it runs
> on the targets must people are developing on (GNU/Linux,Windows),
> it's fine, we're reasonably well covered.

  malloc should be OK,
does it require a header?

Pierre