In message <[EMAIL PROTECTED]> (on 29 October 2004
22:41:32 +0200), [EMAIL PROTECTED] (H.Merijn Brand) wrote:
>On Fri 29 Oct 2004 20:12, Ed Allen Smith <[EMAIL PROTECTED]> wrote:
>>
>> Problem: reports do not appear to include (or to be likely to include from
>> my reading of the current code) all necesary information for
>> diagnosing problems. I suggest that any line containing /\bpanic:\s+/ should
>> be put into the report, such as those in the following:
>
>I think I agree. I don't know why they are filtered. Did you
>have a look in the .log file on the smoke directory?
That's where I'm seeing these; they are not to be found in the mktest.out
file (all it says is "Unable to make anything but miniperl in this
configuration", right after the "Compiler info:" line), and I don't _think_
they are going to come out in the report. It has not finished smoketesting,
however, so I could be wrong.
>> CCCMD = cc -n32 -DPERL_CORE -c -DDEBUGGING -DDEBUGGING -D_BSD_TYPES -D_BSD_TIME
>> -woff 1184,1552 -DLANGUAGE_C -O3 -OPT:Olimit=0:space=ON -O2
>> cc WARNING: -O2 conflicts with -O3; using latter value (-O2)
>> panic: memory wrap.
>> panic: memory wrap.
>> gmake[1]: *** [lib/Config.pm] Error 255
>> gmake: [minitest.prep] Error 2 (ignored)
>> panic: memory wrap.
>> gmake: [minitest] Error 255 (ignored)
>> panic: memory wrap.
>> Unable to make anything but miniperl in this configuration
>>
>> (It would appear from the above, and current testing with -64, that at least
>> in -n32 mode, usemallocwrap does not work on IRIX 6.5; it may work with -64
>
>Can you come up with a patch to the appropriate hint file
>(hints/irix_6.sh) to deal with this?
The easiest way would appear to be to add:
usemallocwrap=${usemallocwrap:-false}
at the end of the file, right under the equivalent statement for
usemymalloc, but that would be for both -n32 and -64 (compiling
using -o32 is not something I would recommend on IRIX 6.5) and would be for
IRIX 6.2+ (6.0 and 6.1 have their own hints); until we know whether
malloc-wrapping works at all on:
A. IRIX 6.2-6.4; and
B. IRIX 6.5 with -64 (or simply doesn't do anything under the test
conditions, but may give a false alarm at a later point)
I am not completely comfortable with this. I could also write a patch that
would only happen with -n32 and IRIX 6.5, but, if problems do happen with
other IRIX versions or with -64, that leaves those people in trouble of
various sorts.
Huh. I just thought of another variation. I should check this out on a
machine that is not 64-bit capable, to see if it displays the error with
-n32. I will attempt to do this this evening, but I do have the concern that
we currently have students using every machine available to do an assigment
due Sunday (midnight), and unfortunately the machines not capable of -64 are
also the ones with the slowest processors, smallest memories, etcetera.
>> (testing not complete, but so far so good), or the test it enables may
>> simply _never_ go off, even if something went wrong, under -64 - I am not a
>> malloc guru (note that Perl's malloc does not function on IRIX, which may be
>> related). I will be testing with -Uusemallocwrap shortly.)
>
>Same here, but I'd prefer second opinions from other people
>with Irix-6.5 access
Strongly agreed - for one thing, it's been a while since I worked with any
(computer) language other than Perl! - plus wanting to know for people with
IRIX 6.2-6.4; I note that Abe has a IRIX 6.2 machine doing smoketesting,
although it seems to be doing 5.9.x, which may not have the problem in
question, or it may be that the problem does not show up on a machine not
capable of -64 (IP22 - especially with R4600 - indicate that it would not be
capable, I believe).
-Allen
--
Allen Smith http://cesario.rutgers.edu/easmith/
February 1, 2003 Space Shuttle Columbia
Ad Astra Per Aspera To The Stars Through Asperity
