This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [patch][+7.3] Fix physname completion regression
- From: Tom Tromey <tromey at redhat dot com>
- To: Jan Kratochvil <jan dot kratochvil at redhat dot com>
- Cc: gdb-patches at sourceware dot org, Keith Seitz <keiths at redhat dot com>
- Date: Mon, 18 Apr 2011 14:53:28 -0600
- Subject: Re: [patch][+7.3] Fix physname completion regression
- References: <20110409203454.GA7430@host1.jankratochvil.net>
>>>>> "Jan" == Jan Kratochvil <jan.kratochvil@redhat.com> writes:
Jan> this is because psymtabs (and even .gdb_index) no longer contain
Jan> parameters type information, therefore the bare names creep in.
Thanks.
Jan> By this fix is clear the completion now requires after physname to
Jan> always expand all the CUs involved. This is in fact even for
Jan> .gdb_index a performance regression against pre-physname
Jan> pre-.gdb_index operation. Maybe .gdb_index could contain all the
Jan> names with their type information - generated with physname by full
Jan> CUs expansion during .gdb_index generation? It is questionable
Jan> whether it makes sense to optimize this case, too many items for
Jan> completion offer are not practically useful to the user anyway.
I am not very concerned about it.
However, if it is a problem, it seems to me that maybe we could
differentiate `break klass::method<TAB>' from `break klass::method(<TAB>',
and let the former search just "partial symbols" while the latter
would require CU expansion
Anyway it looks good to me, thanks.
Tom