Re: [PATCH] malloc/obstack: Merge obstack code back from gnulib

[Will Newton <will dot newton at linaro dot org>]

On 29 March 2014 00:52, Roland McGrath <roland@hack.frob.com> wrote:
> As a general rule, we want each change described fully in the ChangeLog and
> subject to our independent review.  We have to be this anal because we have
> some issues to worry about that gnulib doesn't (and of course vice versa).
> It's certainly worthwhile to mention at the top of the log paragraph that
> you're sync'ing up with gnulib, but the individual changes need to be
> described (and reviewed fully).
>
> These particular changes look mostly quite harmless.  But the principle
> stands.  The ones that are literally cosmetic (changing comments and
> formatting) can be logged trivially (no need to cite every function in the
> normal way, for example).  Adding attributes and such, while also harmless
> in the cases I saw, needs to be logged explicitly.  Renaming functions
> absolutely must be logged--most especially for globals, and that is most
> likely not at all harmless and needs both careful review and your testimony
> that you tested the changes in a libc build including 'make check' (start
> with the quicker 'make check-abi' to confirm or refute my suspicion about
> why the change might be bad before wading through 'make check' and
> identifying all possible regressions).

I tried make check-abi with Paul's most recent set of patches and it
appears to pass successfully on x86_64.

I also did a make check and the failures don't appear to be obviously
related to these changes:

FAIL: elf/tst-dlopen-aout
FAIL: resolv/tst-leaks
Summary of test results:
      2 FAIL
   1702 PASS
    122 XFAIL
      3 XPASS

Is there a specific concern you have about this patch besides the
noted issue of the changelog?

Thanks,

-- 
Will Newton
Toolchain Working Group, Linaro