Re: Patch to dwarf2out.c:gen_block_die

[Mark Mitchell <mark at codesourcery dot com>]

>>>>> "Jason" == Jason Merrill <jason@cygnus.com> writes:

    Jason> This patch fixes building __divdi3 in libgcc; otherwise we
    Jason> assume that none of the blocks in the abstract instance of
    Jason> an inline function are needed, which is wrong.  Mark, does
    Jason> this make sense to you?

I guess.  I'm not a DWARF expert, by any means.

The constraints are these: we're now not calling dwarf2out_begin_block
for all blocks.  (Only for those in the current function that end up
with TREE_ASM_WRITTEN set.)  I'm not sure whether TREE_ASM_WRITTEN is
not deterministic for blocks from other functions.

In general, it should be safe to put out the abstract blocks, as long
as they don't depend on dwarf2out_begin_block being called first for
those blocks.

Also, does the bug occur with plain DWARF as well as DWARF2?  If so,
we might need your fix in dwarfout.c as well.

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com