Remove -Wno-ignored-attributes from fusl build

Remove -Wno-ignored-attributes from fusl build

Many |foo64| symbols are implemented as weak aliases to |foo| via the
LFS64 macro. In a few cases, |foo| itself is also a weak alias to
|__foo|. The semantics of this are such that if |foo| itself is
overridden, |foo64| will still resolve to |__foo|. Clang warns about
this as it is potentially confusing. In these cases, we just write the
weak_alias directly.

R=viettrungluu@chromium.org

Committed: https://chromium.googlesource.com/external/mojo/+/b02a267b65aa63a22315a8497b3bfa61fac57858

[kulakowski, 2016-01-12 20:47:03]

Ditto re openness to alternatives. I don't think the LFS64 macro was buying
much, but if we want the source to preserve it, we could push/pop an override
for the warning in the LFS macro itself, perhaps.

[viettrungluu, 2016-01-12 23:15:26]

LGTM, though I think we should get rid of the LFS64 macro entirely (you don't
have to do that here/now).

(Also, relatedly, which of the ...64() functions are actually specified by
POSIX?)

https://codereview.chromium.org/1579273002/diff/1/fusl/src/stat/statvfs.c
File fusl/src/stat/statvfs.c (right):

https://codereview.chromium.org/1579273002/diff/1/fusl/src/stat/statvfs.c#new...
fusl/src/stat/statvfs.c:61: LFS64(statvfs);
I wonder if we shouldn't just get rid of LFS64 (and presumably also LFS64_2)?

(I wonder if we shouldn't list all the |weak_alias|es at the end of the file or
something. I also wonder why we need, e.g., __statfs -- instead of just calling
it statfs.)

[kulakowski, 2016-01-13 00:35:57]

On 2016/01/12 23:15:26, viettrungluu wrote:
> LGTM, though I think we should get rid of the LFS64 macro entirely (you don't
> have to do that here/now).
> 
> (Also, relatedly, which of the ...64() functions are actually specified by
> POSIX?)
> 
> https://codereview.chromium.org/1579273002/diff/1/fusl/src/stat/statvfs.c
> File fusl/src/stat/statvfs.c (right):
> 
>
https://codereview.chromium.org/1579273002/diff/1/fusl/src/stat/statvfs.c#new...
> fusl/src/stat/statvfs.c:61: LFS64(statvfs);
> I wonder if we shouldn't just get rid of LFS64 (and presumably also LFS64_2)?
> 
> (I wonder if we shouldn't list all the |weak_alias|es at the end of the file
or
> something. I also wonder why we need, e.g., __statfs -- instead of just
calling
> it statfs.)

I *think* all the 64 symbols are linux specific, that they are all implemented
as weak aliases plus a #define foo64 foo in a header. There are some places
where POSIX introduced new APIs but those are spelled differently, e.g. fseeko.

[kulakowski, 2016-01-13 00:49:01]

On 2016/01/13 00:35:57, kulakowski wrote:
> On 2016/01/12 23:15:26, viettrungluu wrote:
> > LGTM, though I think we should get rid of the LFS64 macro entirely (you
don't
> > have to do that here/now).
> > 
> > (Also, relatedly, which of the ...64() functions are actually specified by
> > POSIX?)
> > 
> > https://codereview.chromium.org/1579273002/diff/1/fusl/src/stat/statvfs.c
> > File fusl/src/stat/statvfs.c (right):
> > 
> >
>
https://codereview.chromium.org/1579273002/diff/1/fusl/src/stat/statvfs.c#new...
> > fusl/src/stat/statvfs.c:61: LFS64(statvfs);
> > I wonder if we shouldn't just get rid of LFS64 (and presumably also
LFS64_2)?
> > 
> > (I wonder if we shouldn't list all the |weak_alias|es at the end of the file
> or
> > something. I also wonder why we need, e.g., __statfs -- instead of just
> calling
> > it statfs.)
> 
> I *think* all the 64 symbols are linux specific, that they are all implemented
> as weak aliases plus a #define foo64 foo in a header. There are some places
> where POSIX introduced new APIs but those are spelled differently, e.g.
fseeko.

Yeah so the only symbols for which that is not the pattern are

__fxstat
__fxstatat
__lxstat
__xstat

readdir64_r
readdir_r,

Of these, the xstat-ish ones are Linuxisms
(https://refspecs.linuxfoundation.org/LSB_3.0.0/LSB-PDA/LSB-PDA/baselib-xstat-...),
and the readdir one just chooses not to spell it readdir_r64. So we could nuke
it entirely, it seems.

[kulakowski, 2016-01-13 16:45:41]

Description was changed from

==========
Remove -Wno-ignored-attributes from fusl build

Many |foo64| symbols are implemented as weak aliases to |foo| via the
LFS64 macro. In a few cases, |foo| itself is also a weak alias to
|__foo|. The semantics of this are such that if |foo| itself is
overridden, |foo64| will still resolve to |__foo|. Clang warns about
this as it is potentially confusing. In these cases, we just write the
weak_alias directly.

R=viettrungluu@chromium.org
==========

to

==========
Remove -Wno-ignored-attributes from fusl build

Many |foo64| symbols are implemented as weak aliases to |foo| via the
LFS64 macro. In a few cases, |foo| itself is also a weak alias to
|__foo|. The semantics of this are such that if |foo| itself is
overridden, |foo64| will still resolve to |__foo|. Clang warns about
this as it is potentially confusing. In these cases, we just write the
weak_alias directly.

R=viettrungluu@chromium.org

Committed:
https://chromium.googlesource.com/external/mojo/+/b02a267b65aa63a22315a8497b3...
==========

[kulakowski, 2016-01-13 16:45:42]

Committed patchset #2 (id:20001) manually as
b02a267b65aa63a22315a8497b3bfa61fac57858 (presubmit successful).