Bug 151316 – Lack of sys-cluster/pvm-3.4.5 GCC4 compatibility = segfault for pvm

• Status: RESOLVED FIXED
• Product: Gentoo Linux

[Description Rick Harris 2006-10-14 03:56:47 UTC]

The current gcc4 patch 'pvm-3.4.5-gcc-4.1.patch' is bogus and produces a pvm console binary that segfaults immediately.

Here's the problem showing up in the compile:
../../src/global.c:210: warning: initialization from incompatible pointer type
../../src/global.c:210: warning: excess elements in scalar initializer
../../src/global.c:210: warning: (near initialization for ‘pvmtevdidlist’)
../../src/global.c:211: warning: excess elements in scalar initializer
../../src/global.c:211: warning: (near initialization for ‘pvmtevdidlist’)
../../src/global.c:211: warning: excess elements in scalar initializer
../../src/global.c:211: warning: (near initialization for ‘pvmtevdidlist’)

Here's the problem in the end product:
~ # pvm
Segmentation fault

Attached is a direct replacement for 'pvm-3.4.5-gcc-4.1.patch'

[Comment 1 Rick Harris 2006-10-14 03:59:06 UTC]

Created attachment 99646 [details, diff]
pvm-3.4.5_gcc4.patch

[Comment 2 Stephen J. Turnbull 2006-10-29 01:10:28 UTC]

I'm seeing the same issue, and resolved it (for my local use at least).

I've only tried it on the amd64 platform, but simply reverting the pvm-3.4.5-gcc-4.1.patch works for me.  I don't understand what that patch was trying to accomplish; it doesn't make any sense in standard C!  As for the replacement patch, what is the rationale for moving those declarations out of the headers (which is where they belong IMO)?

Note that it doesn't segfault until after it starts the pvmd3 daemon (note, that's pvmd3, "killall pvm" doesn't work!)

[Comment 3 Tomasz Męciński 2007-03-02 22:37:52 UTC]

(In reply to comment #1)
> Created an attachment (id=99646) [edit]
> pvm-3.4.5_gcc4.patch
> 

Substituting pvm-3.4.5-gcc-4.1.patch with this file helped me on x86; pvm stopped segfaulting.

[Comment 4 Jakub Januszkiewicz 2007-03-08 11:16:43 UTC]

(In reply to comment #3)
> (In reply to comment #1)
> > Created an attachment (id=99646) [edit]
> > pvm-3.4.5_gcc4.patch
> > 
> 
> Substituting pvm-3.4.5-gcc-4.1.patch with this file helped me on x86; pvm
> stopped segfaulting.
> 

Works for me, too.

[Comment 5 Tomasz Jedzierowski 2007-03-08 17:00:00 UTC]

(In reply to comment #3)
> (In reply to comment #1)
> > Created an attachment (id=99646) [edit]
> > pvm-3.4.5_gcc4.patch
> > 
> 
> Substituting pvm-3.4.5-gcc-4.1.patch with this file helped me on x86; pvm
> stopped segfaulting.
> 

I had issues with pvm segfaults despite running /etc/profile, a reboot dealt with it but then i kept getting "libpvm mksocs() connect: Connection refused" errors. I'm running GCC 3.4.6 on x86 but this patch helped out.

[Comment 6 Joshua Hoblitt 2007-12-14 20:51:49 UTC]

The attached patch works for me on amd64/gcc-4.1.2.

[Comment 7 Radoslaw Szkodzinski 2009-05-02 15:16:56 UTC]

Hey, fix this already!

[Comment 8 Marcello Romani 2009-05-12 12:19:23 UTC]

Hi, anybody care to fix this ? Thanks

[Comment 9 Diego Elio Pettenò (RETIRED) 2009-12-06 13:38:41 UTC]

*** Bug 277312 has been marked as a duplicate of this bug. ***

[Comment 10 Andrey Kislyuk (RETIRED) 2009-12-06 19:16:47 UTC]

Taking bug. Unable to initially reproduce issue with pvm-3.4.6. Will take care of pvm and bindings.

[Comment 11 Tomáš Chvátal (RETIRED) 2010-02-17 15:38:05 UTC]

Treecleaned.

[Comment 12 Jeremy Murphy 2010-03-25 12:04:56 UTC]

Did 3.4.6 ever make it to the tree?  I just compiled it without patches using gcc-4.3.4 and the console program did not segfault.  But I guess... is there any demand for it?