This is the mail archive of the
gdb-patches@sources.redhat.com
mailing list for the GDB project.
Re: top-level removal of dejagnu, expect
On Thu, Jun 10, 2004 at 10:14:46AM +1000, Ben Elliston wrote:
>> 1. Have you asked EVERY project that uses the src repository?
>> binutils and gdb aren't the only projects. Why not just remove
>> those directories from the modules file instead? cgf has already
>> mentioned that src/expect has cygwin-specific code in it.
>
>I have asked the major players and got overwhelming support. Removing
>the directories from the modules file is broken, as you know. It
>should be possible to check those files out if they were included in a
>past tag (which they were).
>
>As for Expect, the version in src/expect has not been updated since
>sourceware was instated last century :-) There are much newer versions
>which carry fixes for some of the problems that are being reported. I
>am pushing all patches made since the import to Don Libes and do not
>intend to drop any. I found a couple of patches related to Cygwin.
>Many of them were removed by Chris Faylor in 2001, however.
I think DJ removed a few, too. The basic problem was that expect was
trying to invent ptys on Windows. Since cygwin already has ptys, a lot
of code could go away.
>> 2. Our internal repository has a customized dejagnu; it would be
>> convenient for us if support for an in-tree dejagnu were
>> maintained. Likewise for other developers who may want to (or need
>> to) customize dejagnu for new ports in conjunction with porting the
>> tools.
>
>I stated at the GCC Summit that, as a DejaGnu maintainer, I will be
>responsive to urgent bug fixes. I appreciate that. Those fixes can
>be made available via the Savannah CVS server, or I can push out
>maintenance releases as required. No one felt this was a problem.
If it was anyone else but you Ben, I might have reservations. I know
that you will be enthusiastically responsive to any problems.
>> So, I would prefer that support for an in-tree dejagnu were
>> retained. Even if dejagnu and expect were totally wiped from the
>> src repository, the toplevel changes would not be needed to support
>
>An in-tree dejagnu is a pain to maintain. The last six months have
>demonstrated this. I merged the two dejagnu trees at the end of
>January and since then, have struggled to keep them in sync.
That's why I can't object too strongly to this. I don't like mirroring
other people's repositories on sourceware since it just guarantees
drift. Readline suffers similarly.
So, even if it causes a little pain now, the net gain should be worth
it.
cgf