[vta,vta4.3,trunk?] defer creation of DECLs for dwarf indirect constants

[Alexandre Oliva]

On Sep 10, 2008, Alexandre Oliva <aoliva@redhat.com> wrote:

> This is another difference detected by -fcompare-debug, that *may*
> cause codegen differences with or without -g.  At the end of
> compilation of a function that requires constants to be placed in
> memory for debug info, we create new decls.  These new decls cause
> decls in subsequent functions to vary depending on whether debug
> information is generated.  Varying decl uids are known to cause
> codegen differences in some cases, although I don't have any such
> testcase.  They were enough to change anonymous temporary D.<number>
> names in dumps generated for -fcompare-debug, though, which was enough
> of a reason for me to fix it.

> The fix, which should also reduce total memory use during compilation,
> was to delay the creation of DECLs for the constants to the end of
> compilation, when we actually need them to emit debug info.  Until
> then, we hold on to the identifier we'd have used for the DECL.

> Ok for trunk?

Ping?

-------------- next part --------------
A non-text attachment was scrubbed...
Name: vta-dwarf2asm-defer-decl-uid.patch
Type: text/x-patch
Size: 3059 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20090601/645667ce/attachment.bin>
-------------- next part --------------

-- 
Alexandre Oliva, freedom fighter    http://FSFLA.org/~lxoliva/
You must be the change you wish to see in the world. -- Gandhi
Be Free! -- http://FSFLA.org/   FSF Latin America board member
Free Software Evangelist      Red Hat Brazil Compiler Engineer