This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH] fix remote host test failures in testsuite
- From: Tom Tromey <tromey at redhat dot com>
- To: Hui Zhu <teawater at gmail dot com>
- Cc: gdb-patches ml <gdb-patches at sourceware dot org>, Sandra Loosemore <sandra at codesourcery dot com>, Hui Zhu <hui at codesourcery dot com>, Pedro Alves <palves at redhat dot com>
- Date: Thu, 18 Jul 2013 12:27:38 -0600
- Subject: Re: [PATCH] fix remote host test failures in testsuite
- References: <CANFwon18_v9PtqmA8yw_CesvyqzHiwn1vqNViUAKfb0oLXsg7Q at mail dot gmail dot com>
>>>>> "Hui" == Hui Zhu <teawater@gmail.com> writes:
Hui> This patch will make GDB testsuite support remote host test better.
Hui> It updates gdb.exp's function build_id_debug_filename_get and
Hui> gdb_gnu_strip_debug to support remote host.
Thanks.
Hui> It updates info-fun.exp, solib-display.exp and solib-nodir.exp to let
Hui> them skip test if remote host because they base on share libary but I
Hui> didn't find good way to handle it in remote host.
That's unfortunate but it does seem a bit tricky. Maybe it could be
done by making the .so in "." when using a remote host.
That would not be parallel-safe, but I expect we'll probably disable
parallelization in this case anyway.
Alternatively, can we make a parallel directory structure on the remote
host? That was actually my plan for the parallelization approach... not
sure it works though. If it does it seems like this would help remove
differences between the two modes, which would be good.
Hui> --- a/gdb/testsuite/lib/gdb.exp
Hui> +++ b/gdb/testsuite/lib/gdb.exp
[...]
Hui> + set local_tmp [remote_upload host "$tmp"]
I think this changes behavior when not doing remote-host testing.
The old style was not to care; but I think this hurts parallelization,
so I do care.
I wouldn't mind a wrapper proc to make it all easy though.
Hui> + remote_file delete "$host_dest"
Hui> + remote_file delete "$host_debug_file"
Hui> + remote_file delete "$stripped_file"
I think remote_file takes a first argument like "host".
So these calls, I think, do nothing.
Tom