From mboxrd@z Thu Jan  1 00:00:00 1970
From: Andrew Cagney <cagney@gnu.org>
To: Daniel Jacobowitz <drow@false.org>
Cc: Paul Hilfinger <hilfingr@gnat.com>, jimb@redhat.com, gdb-patches@sources.redhat.com
Subject: Re: [RFA] Introduce notion of "search name"
Date: Wed, 12 May 2004 15:11:00 -0000
Message-id: <40A23E97.1020906@gnu.org>
References: <20040429103706.4793BF2BFF@nile.gnat.com> <20040429211458.GB27523@nevyn.them.org> <vt2n04umj8b.fsf@zenia.home> <20040430084538.ECDE1F2E1C@nile.gnat.com> <20040430134955.GA15786@nevyn.them.org> <20040503084937.439F4F2C0A@nile.gnat.com> <20040511194843.GA15952@nevyn.them.org> <20040512105959.806E6F2DE4@nile.gnat.com> <20040512132708.GA25122@nevyn.them.org> <40A2313E.1080100@gnu.org> <20040512142254.GA26822@nevyn.them.org>
X-SW-Source: 2004-05/msg00369.html

A quick grep of GDB shows that we're already doing:

$ grep -e SYMBOL_PRINT_NAME -e SYMBOL_CPLUS_DEMANGLED_NAME -e 
DEPRECATED_SYMBOL_NAME -e SYMBOL_NATURAL_NAME -e SYMBOL_SOURCE_NAME -e 
SYMBOL_LINKAGE_NAME *.c  | grep -v -e ^sym -e syms.c: -e read.c: | cut 
-d: -f1 | sort -u
ada-lang.c
ada-typeprint.c
ax-gdb.c
blockframe.c
breakpoint.c
buildsym.c
c-valprint.c
cp-namespace.c
cp-support.c
cp-valprint.c
dictionary.c
dwarf2loc.c
expprint.c
f-valprint.c
findvar.c
glibc-tdep.c
gnu-v2-abi.c
hppa-hpux-tdep.c
i386-tdep.c
infcall.c
infcmd.c
jv-lang.c
linespec.c
maint.c
objc-lang.c
objfiles.c
p-valprint.c
ppc-sysv-tdep.c
printcmd.c
rs6000-tdep.c
solib-frv.c
somsolib.c
stack.c
tracepoint.c
typeprint.c
valops.c

I.e., through out GDB there is the assumption that symtab internally 
uses simple strings.

We shouldn't.  Recalling that one of the underlying problems here was 
the need to construct search names on the fly using more complext debug 
info, we should instead have interfaces such as:

symtab_name_put (symtab, ui_file):
Write the printable name to the specified output.
symtab_demanged_cmp (block, symtab, symtab):
Compare to symbols returning their relative position.
that completly abastract symbol name lifetimes (you never know if it is 
defined).

By doing that we better specify the symtab interface; clarify the 
clients requirements; and free the symtab code of specific 
implementation assumptions.

We can also finally answer questions such as:
- how often is the printed name used
- how often is a symbol comparison performed
and use that to adjust the internal representation so that it better 
meets real needs

Andrew