Re: maxima: The SGC segfault recovery test failed with memprotect_bad_fault_address, SGC disabled

[Kei Kebreau]

Alex Vong <address@hidden> writes:

> Kei Kebreau <address@hidden> writes:
>
>> Alex Vong <address@hidden> writes:
>>
>>> Hello all,
>>>
>>> When I start maxima, I get the following warning message:
>>>
>>> The SGC segfault recovery test failed with
>>> memprotect_bad_fault_address, SGC disabled
>>> The SGC segfault recovery test failed with
>>> memprotect_bad_fault_address, SGC disabled
>>> Maxima 5.41.0 http://maxima.sourceforge.net
>>> using Lisp GNU Common Lisp (GCL) GCL 2.6.12
>>> Distributed under the GNU Public License. See the file COPYING.
>>> Dedicated to the memory of William Schelter.
>>> The function bug_report() provides bug reporting information.
>>>
>>> Actually I encountered this problem quite a while ago after an update,
>>> but I didn't report it since it was not fetal.
>>
>> When did you first notice this issue?
>>
> I am afraid I cannot remember exactly when...
>
> I installed wxmaxima long ago and things worked fine for a long
> time. One day, after updating wxmaxima, it showed the above warning and
> started to crash nondeterministically when I typed things into the
> message box. The crash was fixed quickly by another update, but the
> warning persisted. 
>
>>>
>>> Does anyone knows what's going on?
>>>
>>> Cheers,
>>> Alex
>>
>> Hello Alex,
>>
>> This is my first time running into this issue. While I do get this
>> warning message using the Maxima substitute from hydra.gnu.org, when I
>> build Maxima 5.41.0 from source locally, this warning message does not
>> appear. Is this the case for you as well?
>>
> Same for me. I set --prefix during configure and install it
> somewhere. It starts without the warning.
>
>> The message comes from the source code of GNU Common Lisp, but the GCL
>> binary from hydra.gnu.org doesn't give this error when I run it. It
>> seems like this comes down a bug in GCL set off by Maxima, a bug in
>> Maxima, a reproducibility issue, or some combination of all of
>> these. I'll be investigating this further.
>>
>> Thank you,
>> Kei
>
> Thanks for looking at it!

I turns out that the issue is with stratified garbage collection in
GCL. The substitute from hydra.gnu.org does in fact give me the error
when I run "(si::sgc-on t)" in the REPL. I'm preparing a patch to update
GCL to the version used in Debian stable (GCL git tag "pathnames1.13") and
we'll see if changing that definition on the master branch will do anything.

signature.asc
Description: PGP signature