Re: [PATCH 0/5] Rename a few internal symbols to avoid collisions

[Pedro Alves <palves at redhat dot com>]

On 02/17/2017 02:16 AM, Pedro Alves wrote:
> On 02/17/2017 01:06 AM, Alan Modra wrote:
>> On Fri, Feb 17, 2017 at 12:17:55AM +0000, Pedro Alves wrote:
>>>   bfd: Rename read_{signed,unsigned}_leb128, safe_read_leb128
>>>   bfd: Rename real_{ftell, fseek, fopen}
>>>   bfd: Rename warn_deprecated
>>>   bfd: Rename bsd_write_armap and coff_write_armap
>>>   bfd: Rename Chunk and S3Forced
>>
>> OK, thanks.
>>
> 
> Thanks, I pushed it, and the buildbot complained that I broke
> the build...  There are a few instances of safe_read_leb128
> in bfd/dwarf2.c that I didn't adjust.  I was puzzled,
> since my tree builds cleanly.  But removing the .o files and
> rebuilding the file triggers it.  Looks like I hit some
> stale dependency.  :-/  Fixing...

Ah, no, false alarm, actually.  master is fine.  It's simply that
there's one hunk that was accidentally left in patch #2 that should
have been in patch #1:

--- a/bfd/dwarf2.c
+++ b/bfd/dwarf2.c
@@ -906,7 +906,8 @@ read_abbrevs (bfd *abfd, bfd_uint64_t offset, struct dwarf2_debug *stash)
 
   abbrev_ptr = stash->dwarf_abbrev_buffer + offset;
   abbrev_end = stash->dwarf_abbrev_buffer + stash->dwarf_abbrev_size;
-  abbrev_number = safe_read_leb128 (abfd, abbrev_ptr, &bytes_read, FALSE, abbrev_end);
+  abbrev_number = _bfd_safe_read_leb128 (abfd, abbrev_ptr, &bytes_read,
+                                        FALSE, abbrev_end);
   abbrev_ptr += bytes_read;
 

So all the built slaves will fail to build patch #1.  They should go back
to normal automatically once they move past that commit.

Thanks,
Pedro Alves