This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Changes introduced with b-i-b merge
- From: Loren James Rittle <rittle at latour dot rsch dot comm dot mot dot com>
- To: gcc at gcc dot gnu dot org
- Date: Wed, 18 Dec 2002 16:09:41 -0600 (CST)
- Subject: Changes introduced with b-i-b merge
- Reply-to: rittle at labs dot mot dot com
I am fairly caught up on mailing list traffic but it appears that
various conventions changed with the b-i-b merge. To rebuild an
object tree with a trivial source change/minor update, I used to say
at top-level: ``gmake Makefile && gmake -C gcc quickstrap && gmake''
If I also wanted to force a library to be rebuilt from scratch with
the new compiler, I merely removed <target-triple>/<library> beforehand.
If I ran the above sequence, then 99/100 times it would do the right
thing. Most importantly, it would not do a lot of extra work just to
recompile a minor change. If I understand properly, I now use:
``gmake'' at top-level to get the same thing as the above command line.
The question is: Why does it now do so much work even when no files
have changed? For a start, it appears to reconfigure every directory.
For some reason, this gratuitous reconfigure is forcing, e.g., a lot
of libjava to be rebuilt (thank Supreme Being that it is not all of
it) and all of libf2c even though nothing changed.
I.e. this is a very annoying change. Before b-i-b merge, top-level
gmake when no files had been changed required about 20-40 seconds on
this machine (most of the time was spent running the libjava
Makefile). Now, it takes ~30 minutes when NOTHING changed in my
source tree. Obviously, I'm not advocating that we deinstall this
merge for such a trivial reason as developer productivity but could
the people responsible please figure out how to fix this.
For the record, this is on mainline updated ``Wed Dec 18 22:04:18 GMT 2002''
however I am noting rebuild times in a tree fully bootstrapped last night.
Regards,
Loren