[PATCH] archives require a symbol table on Solaris, even if empty

[Danek Duvall via llvm-commits]

Danek Duvall wrote:

> Rafael Avila de Espindola wrote:
> 
> > Danek Duvall <danek.duvall at oracle.com> writes:
> > 
> > > Rafael Avila de Espindola wrote:
> > >
> > >> Danek Duvall <danek.duvall at oracle.com> writes:
> > >> > I've attached a new patch with an updated comment, a testcase, and an
> > >> > archive, with a git-style patch.  Let me know if there's a better way to do
> > >> > this.
> > >> 
> > >> Since the archive format is basically text, I think you can just use
> > >> FileCheck to test that we create a symbol table on an empty file. Take a
> > >> look at llvm/test/Object/archive-format.test. In fact, don't you have to
> > >> update that test with your patch?
> > >
> > > I don't -- the problem only appears when there's an actual object file in
> > > the archive, not when the archive is empty or contains only text files.
> > >
> > > So I figured out how to use yaml2obj to create a minimal object file.  I
> > > couldn't figure out how to get FileCheck to deal with null characters, so I
> > > used cat -v instead of cat.  Perhaps that last check line isn't strictly
> > > necessary, since the previous line had a 4 in place of the 8 that should be
> > > there, but it's a bit more clear what's going on this way.  I'm happy to
> > > change it, if you don't feel it's necessary or cat -v isn't portable enough.
> > >
> > > I also moved the test into archive-format.test, since it seemed like a
> > > reasonable place for it.
> > 
> > I would probably avoid "cat -v" if not strictly required. The patch
> > looks good with that change.
> 
> I couldn't figure out how to make it work without cat -v, other than
> dropping that last checked line of output.
> 
> > Do you need me to commit it?
> 
> Yes, please.

Is there anything else you need from me on this?  Do you want me to re-send
a patch without the cat -v?

Thanks,
Danek