Re: [RFA 2/4] dwarf2_physname

[Tom Tromey <tromey at redhat dot com>]

>>>>> "Daniel" == Daniel Jacobowitz <drow@false.org> writes:

Keith> Some warnings are in order. First, this is probably going to really
Keith> slow down large C++ applications, because we no longer use
Keith> DW_AT_MIPS_linkage_name as generated from the compiler. We
Keith> essentially build this name during DIE reading.

Daniel> We need to quantify the impact on a couple of C++ code bases, I think.

I think it is worth noting that this problem should be mostly solved by
some work I'm in the middle of.  This work involves making psymtabs
optional (it is left up to the debuginfo reader to decide), and then
making use of various DWARF indices (we have a new section but debate
about this is ongoing) to speed up the DWARF case.  The upshot of this
is that we'll generally only read CUs on demand, and we won't do the
initial scan we currently do for psymtabs.

This work won't help with existing code, though, because it relies on
indices we don't have access to (even if we use the standard DWARF ones,
those are unusable with older versions of GCC).

If you're interested in this, it is on archer-tromey-optional-psymtab.
You'll also want the compiler and readelf patches from:

    http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41130

This is a work in progress.

Tom