[darcs-users] Re: Segfaults on Solaris 8/SPARC darcs 1.0.2

Jason Dagit dagit at cs.orst.edu
Tue Feb 15 23:16:19 UTC 2005



On Wed, 16 Feb 2005, [ISO-8859-1] Anders Höckersten wrote:

> It wasn't clear to me from your description: Did you try compiling darcs in 
> EXACTLY the same way (i.e with libcurses and libcurl enabled) as before, but 
> using -fvia-c instead of -O? If that works, I would have to say your 
> suggestion makes (some) sense. I still think the problem has to do with your 
> particular installation, NOT with Solaris 8 in general. Unfortunately I only 
> have access to Solaris 9 machines, so I am unable to verify this.

Yes, I compiled the same way with the only differences being -fvia-C and 
-O, and when I use -O I get segfaults.  You could be right about it not 
affecting Solaris 8 in general, but I also don't know how to verify this 
as I don't have access to a clean standard installation of Solaris 8.

> According to the ghc manual, -O implies -fvia-c. Whilst -O also enables a few 
> other optimizations, none of them "should" be optimizations that break things 
> on Solaris 8 but not on e.g Solaris 9. I think it is more likely you have a 
> bad combination of compilers and/or libraries that triggers some bug.

Yes, hard to say.

> You might want to try something that uses compressed patches, since someone 
> reported problems with that on Solaris. It seems to work for me though. A 
> simple testcase should be (someone more skilled in using darcs should correct 
> me if this does not test compressed patches, I am still a beginner):
> darcs init
> touch foo
> darcs add foo
> darcs record --compressed

I did the following:
[03:09pm][dagit at ragnarok~/tmp]
$ darcs init
[03:09pm][dagit at ragnarok~/tmp]
$ darcs add test
[03:09pm][dagit at ragnarok~/tmp]
$ darcs record
addfile ./test
Shall I record this patch? (1/1) [ynWsfqadjk], or ? for help: y
What is the patch name? test
Do you want to add a long comment? [yn] n

darcs failed:  Error opening file 
_darcs/patches/20050215230942-3c84d-50d9de7a08e841073d03d181440137b115cd7a3d.gz-0

That made me realize that it was failing on a compressed patch, so I tried 
with --dont-compress and it was able to record the patch.  So it would 
appear that I'm having issues with compressed patches.

Because of this I will continue to use darcs from the redhat machine I 
found, but this way is sub-optimal because it relies on my bash_profile to 
change the value of $HOME.

If anyone has more ideas of things to try, I'm game.

Jason


More information about the darcs-users mailing list