unicos compile problem with 5.005_56 #1119
Comments
From jsalinas@sgi.comIt seems I have run into yet another snag trying to install Last time I had a problem with perl5.005.02 it was because The problem I have now seems centered around the same When I run configure using the defaults and use the Your system wants malloc to return 'void *', it would seem. When the make runs I receive the following message: The following tests fail: Make: " cd t && (rm -f perl; /bin/ln -s ../miniperl perl) \ Then the make core dumps. The debug print shows: ################################################################################ ***** START OF SYMBOLIC DUMP ***** LIST OF PROCESS STATES CPUs 35593: Signal SIGERR <Error exit> DISPLAYING CPUs 35593: Signal SIGERR <Error exit> Signal SIGERR in routine Perl_filter_read at address 0p73511a ***** END OF SYMBOLIC DUMP ***** ################################################################################ I tried removing the miniperl and compiling it by hand: then with miniperl I could print out a version number as 5.005_63 At this point I tried removed the perl directory and redid the configure that miniperl was so broken that if I tried to print out I tried moving the term.t file to /t/lib and I am compiling Any suggestions on where I should go to next? Any suggestions thanks, |
From [Unknown Contact. See original ticket]John Salinas writes:
Perl failed to convince sbrk() to return something on 0x800-byte -D optimize=-g and then run miniperl -Dm -e0 ? Ilya |
From [Unknown Contact. See original ticket]Using the unicos supplied malloc and the -D optimize=-g flag If I use the perl supplied malloc with the -D optimize=-g flag I successfully compiled both of the above mentioned versions thanks, Make Messages With unicos malloc and the -D optimize=-g flag on /perldebug.tmp /perldelta.tmp perldiag.tmp perltrap.tmp perlhpux.tmp With perl malloc and -D optimize=-g flag nomad% make Old E-mail Message
|
From [Unknown Contact. See original ticket]On Thu, Feb 03, 2000 at 03:15:20PM -0600, John Salinas wrote:
You still did not report results of running requested program... Ilya |
From [Unknown Contact. See original ticket]nomad% /opt_ieee/ctl/bin/cc -D optimize=-g -o miniperl miniperlmain.o opmini.o libperl.a -lm -lc -lcrypt -lnsl I am assuming though that was because I was still uing the ./miniperl -w -Ilib -MExporter -e 0 || make minitest I also no longer have to move term.t and edit toke.c by One other very minor thing it woudl be nice to have line thanks,
|
From [Unknown Contact. See original ticket]On Thu, Feb 03, 2000 at 03:33:04PM -0600, John Salinas wrote:
I somewhat confused: you did not put -D optimize=-g on the Configure command line (after renaming config.sh)? (With
This should not happen if you did as I said. Ilya |
From [Unknown Contact. See original ticket]I did not rename config.sh here were the steps I went This would work for unicos malloc and not the perl malloc, the After re-reading your e-mails I tried this: On the make it blew up on the malloc again: CC-9 cc: WARNING File = xsutils.c, Line = 26 * bundled lib/*.pm files is in a version-specific directory, `sh cflags libperl.a globals.o` globals.c But I still did try the following: I have not looked at the core yet but I am guessing Either way unless I misunderstand your instructions it thanks,
|
From [Unknown Contact. See original ticket]On Thu, Feb 03, 2000 at 05:57:03PM -0600, John Salinas wrote:
I did not see sh Configure -D optimize=-g Ilya |
From [Unknown Contact. See original ticket]It is there you did not read far enough down I will I did not rename config.sh here were the steps I went This would work for unicos malloc and not the perl malloc, the After re-reading your e-mails I tried this: -------------------------> On the make it blew up on the malloc again: CC-9 cc: WARNING File = xsutils.c, Line = 26 * bundled lib/*.pm files is in a version-specific directory, `sh cflags libperl.a globals.o` globals.c But I still did try the following: I have not looked at the core yet but I am guessing Either way unless I misunderstand your instructions it thanks,
|
From [Unknown Contact. See original ticket]On Thu, Feb 03, 2000 at 06:34:20PM -0600, John Salinas wrote:
I did and it is not. I think you did an analogue of -D optimize="-D optimize=-g"
Thanks.
Folks, do we want to change this? I expect M$ compilers be as stupid
Hmm, it is reporting different errors now... Will think.
Perl malloc can handle the 64 bit system or word addressing. Ilya |
From [Unknown Contact. See original ticket]On Thu, Feb 03, 2000 at 06:34:20PM -0600, John Salinas wrote:
Stupid me! Of course one would not be able to debug such problems
Looks like bugs in debugging code... And the debugging code looks I made some minor adjustment to the debuggging code. Please retry with Inline Patch--- ./malloc.c~ Thu Jan 27 17:53:27 2000
+++ ./malloc.c Fri Feb 4 17:42:12 2000
@@ -1034,8 +1034,14 @@ Perl_malloc(register size_t nbytes)
if ((PTR2UV(p)) & (MEM_ALIGNBYTES - 1)) {
dTHXo;
PerlIO_printf(PerlIO_stderr(),
- "Corrupt malloc ptr 0x%lx at 0x%"UVxf"\n",
- (unsigned long)*((int*)p),PTR2UV(p));
+ "Unaligned pointer in the free chain 0x%"UVxf"\n",
+ PTR2UV(p));
+ }
+ if ((PTR2UV(p->ov_next)) & (MEM_ALIGNBYTES - 1)) {
+ dTHXo;
+ PerlIO_printf(PerlIO_stderr(),
+ "Unaligned `next' pointer in the free chain 0x"UVxf" at 0x%"UVxf"\n",
+ PTR2UV(p->ov_next),PTR2UV(p));
}
#endif
nextf[bucket] = p->ov_next;
Ilya P.S. Whoever edited malloc.c make it unusable outside of Perl... Sigh... |
From [Unknown Contact. See original ticket]Changing the debug print definately made a difference. I was short version: I did not try to put the PL_debug |= 128 in miniperlmain.c thanks, ################################################################################ @@ -1034,8 +1034,14 @@ Perl_malloc(register size_t nbytes) ################################################################################ nomad% rm -f config.sh Policy.sh ################################################################################ nomad% ./miniperl -Dm -e0 ################################################################################ ***** START OF SYMBOLIC DUMP ***** LIST OF PROCESS STATES CPUs 42864: Signal SIGABRT <Abort> DISPLAYING CPUs 42864: Signal SIGABRT <Abort> Signal SIGABRT in routine _lwp_killm at address 0p1075401c ***** END OF SYMBOLIC DUMP ***** ################################################################################
|
From [Unknown Contact. See original ticket]John Salinas writes:
Thank you for persisting on this. I was afraid that you gave up...
I'm puzzled. After the patch there is no code which would emit Ilya |
From [Unknown Contact. See original ticket]John Salinas writes:
Btw, you did not send us the results of running ./myconfig. I suspect that you got size_of_ptr being 8, but size_of_uv being 4. [Unfortunately, Configure maintainers did not make it easy to deduce Perl (including Perl's malloc) assumes that IV is large enough to keep Ilya |
From @jhiIlya Zakharevich writes:
Yes, we did. Configure (config.sh, config.h, %Config) has (I doubt very much uvsize being 4, in Crays all but chars are 8 bytes)
-- |
From [Unknown Contact. See original ticket]Okay here is the myconfig output: nomad% ./myconfig As well as some liens from the config.sh uvoformat='"o"' I tried to compare the perl malloc.c to the thanks, |
From [Unknown Contact. See original ticket]On Fri, Feb 11, 2000 at 04:23:06PM -0600, John Salinas wrote:
Looks very reasonable indeed... Ilya |
From [Unknown Contact. See original ticket]On Sat, Feb 12, 2000 at 12:17:26AM +0200, Jarkko Hietaniemi wrote:
Oh, thank you a lot, and sorry for not keeping in sync with this stuff... |
From [Unknown Contact. See original ticket]Any thoughts about this now I was willing to try something thanks, Forwarded message:
|
|
arthur@contiller.se - Status changed from 'resolved' to 'open' |
|
arthur@contiller.se - Status changed from 'open' to 'resolved' |
Migrated from rt.perl.org#2077 (status was 'resolved')
Searchable as RT2077$
The text was updated successfully, but these errors were encountered: