Fix clang compilation errors in libelf.

Fix clang compilation errors in libelf.

Remove use of VLA inside a union/struct.  Replace nested functions with
local static functions.

Fixes:
  third_party/elfutils/src/libelf/elf32_updatefile.c:294:4: error: function definition is not allowed here
  third_party/elfutils/src/libelf/elf32_updatefile.c:328:7: error: implicit declaration of function 'fill_mmap' is invalid in C99 [-Werror,-Wimplicit-function-declaration]

  third_party/elfutils/src/libelf/elf_getarsym.c:201:15: error: fields must have a constant size: 'variable length array in structure' extension will never be supported
  third_party/elfutils/src/libelf/elf_getarsym.c:202:15: error: fields must have a constant size: 'variable length array in structure' extension will never be supported

  third_party/elfutils/src/libelf/elf_begin.c:1015:3: error: function definition is not allowed here
  third_party/elfutils/src/libelf/elf_begin.c:1047:11: error: implicit declaration of function 'lock_dup_elf' is invalid in C99 [-Werror,-Wimplicit-function-declaration]

BUG=385553

[rmcilroy, 2014-06-27 15:24:26]

Please add a patch file which can be used to re-apply this change, and update
README.chromium to reference this patch

https://codereview.appspot.com/106240043/diff/1/libelf/elf32_updatefile.c
File libelf/elf32_updatefile.c (right):

https://codereview.appspot.com/106240043/diff/1/libelf/elf32_updatefile.c#new...
libelf/elf32_updatefile.c:389: scn_start, dl->data.d.d_off);
I think this should be: 
fill_mmap (last_position, shdr_start, shdr_end,
scn_start, 0);

No?

[simonb1, 2014-06-27 15:52:57]

Fix uploaded in patch set 2.  Ptal, thanks.

https://codereview.appspot.com/106240043/diff/1/libelf/elf32_updatefile.c
File libelf/elf32_updatefile.c (right):

https://codereview.appspot.com/106240043/diff/1/libelf/elf32_updatefile.c#new...
libelf/elf32_updatefile.c:389: scn_start, dl->data.d.d_off);
On 2014/06/27 15:24:26, rmcilroy wrote:
> I think this should be: 
> fill_mmap (last_position, shdr_start, shdr_end,
> scn_start, 0);
> 
> No?

Nice catch, thanks.  Done.

[rmcilroy, 2014-06-27 16:30:16]

lgtm.

[thakis, 2015-05-05 15:56:08]

was this sent upstream in some way?

[simonb1, 2015-05-06 14:38:35]

On 2015/05/05 15:56:08, thakis wrote:
> was this sent upstream in some way?

No.  It only addresses the needed libelf (and even then only for the required
functions inside it), and so is not a full elfutils patch.  Elfutils contains
other libraries.  An elfutils patch suitable for acceptance upstream would
require clang fixes to all elfutils libraries.

Prior discussion on this topic:
https://code.google.com/p/chromium/issues/detail?id=217722#c5

"nested functions tend to be the worse thing to try and make work under clang. 
upstream packages aren't generally interested in reworking the code (because
usually it's embedded pretty far), and clang isn't interested in adding support
for the feature."

[thakis, 2015-05-07 19:41:27]

On 2015/05/06 14:38:35, simonb1 wrote:
> On 2015/05/05 15:56:08, thakis wrote:
> > was this sent upstream in some way?
> 
> No.  It only addresses the needed libelf (and even then only for the required
> functions inside it), and so is not a full elfutils patch.  Elfutils contains
> other libraries.  An elfutils patch suitable for acceptance upstream would
> require clang fixes to all elfutils libraries.
> 
> Prior discussion on this topic:
> https://code.google.com/p/chromium/issues/detail?id=217722#c5
> 
> "nested functions tend to be the worse thing to try and make work under clang.

> upstream packages aren't generally interested in reworking the code (because
> usually it's embedded pretty far), and clang isn't interested in adding
support
> for the feature."

Have you asked upstream if they'd be interested in the patch?