This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [patchv2] Support .dwp with the name of symlinked binary file
- From: Doug Evans <dje at google dot com>
- To: Jan Kratochvil <jan dot kratochvil at redhat dot com>
- Cc: gdb-patches <gdb-patches at sourceware dot org>
- Date: Mon, 16 Sep 2013 16:30:23 -0700
- Subject: Re: [patchv2] Support .dwp with the name of symlinked binary file
- Authentication-results: sourceware.org; auth=none
- References: <20130905131839 dot GA8618 at host2 dot jankratochvil dot net> <CADPb22RKV2hHQXHdy4f0vjxhnfjDVA1AzHUpwNh06xu+NejZWw at mail dot gmail dot com> <20130916190151 dot GA24936 at host2 dot jankratochvil dot net>
Jan Kratochvil writes:
> ISTM people want the directories normalized against any "../".
True.
OTOH, gdb can't know if/when there is an formal abstraction being implemented in symlinks (and thus when to terminate its realpath'ing). I can imagine providing some means to handle this, but we don't have to add it until there's a compelling need.
> IIUC you were proposing thin 'struct gdb_bfd *' wrapper around 'bfd *'.
It was a possibility to explore, nothing concrete.
> In the patchset above I have found we do not IMO need non-canonical filename
> from 'bfd *', it is enough to have non-canonical filename from 'objfile *'.
Sure.
> > And if we record the original file name we can always
> > get the realpath name,
>
> Calling gdb_realpath() dynamically is very expensive.
I wasn't suggesting that (in and of itself). :-)
[modulo, if one were to go in that direction one would implement the realpath cache, and then one might decide whether to still head in that direction]