irclog2html for uclibc on 2002.08.24

00:00:52mjn3ok... trying again.  ;-)
00:02:19anderseeOk.  Everything is in now.  Trying again on powerpc
00:06:25anderseecool.  powerpc seems to be working perfectly once again.
00:09:08mjn3building python
00:11:55anderseetesting arm now...
00:12:55mjn3hmm... i'm seeing a new failure with python.  trying without threads.
00:13:18andersee:-(
00:15:48mjn3trying to track it down
00:16:46anderseearm seems to be compiling fine.
00:17:09anderseeStarting a compile on mips now as well (that one will take a while...)
00:17:57mjn3i've come to the conclusion that i hate mips errnos...
00:18:47anderseeyeah
00:19:02mjn3rebuilding python with thread-less uClibc.  configuring/building perl too
00:19:22anderseeThey apparently smoked a bunch of crack before picking errno values...
00:20:37tzangerha
00:22:54anderseeFound a mips specific types.h bug.  Fixing.
00:23:43mjn3andersee: dircache test is now failing in python
00:23:50mjn3checking details
00:27:10mjn3andersee: false alarm.  don't know what was happening, but dircache is now passing.
00:27:47mjn3andersee: the perl tests (except for sysvipc) are passing
00:31:02anderseecool
00:31:07mjn3andersee: python tests pass with threadless uclibc.  trying threads again
00:31:48anderseeI had to make a couple of changes so mips would compile cleanly.  Still compiling but all looks fine now.
00:32:38anderseedid you do a 'make clean; make' on uClibc.  Those type changes are pretty fundamental..  could explain wierdness...
00:32:50mjn3sure did
00:37:30anderseeArm compile finished and works fine.
00:52:59mjn3andersee: ok.  things seem to be passing.  for some reason my fake librt.a isn't disabling some problems when threads are enabled.
00:53:13mjn3andersee: also, the glibc stdio tests passed.
00:54:16mjn3andersee: i need to figure out why librt is screwing up some of the tests when threads are on.
00:55:03tzangerhttp://www.fatport.com/products/fatpoint_OEM    <-- why would they have gone for an old 80386 when they could have likely had a nice fast ARM processor to do this for cheaper?
00:58:47anderseemjn3: mips compile finished and works fine
00:59:30mjn3andersee: good.  the python tests are passing, except for a couple where librt is screwing things up.  even putting my fake one in, it is still trying to load glibc;s
01:00:19mjn3andersee: erik, i'm shutting down for tonight.  i'll look at the python stuff some more in the morning.  but the perl tests are passing so your changes are probably ok
01:01:59anderseecool bean
01:02:02anderseenight then
01:02:49mjn3night
09:50:55bug1ls sayx busybox is 13800 Bytes, size says (text + data + bss) == 12190, whats in the other 1610 Bytes ?
13:31:21anderseemorning
13:31:26mjn3morning.
13:31:51mjn3just about to start looking at the python librt stuff problems
13:32:01anderseeI've still not hit that pesky shm bug...
13:32:12anderseeGuess I should do that today
13:32:34mjn3that would be good... i want to get the strerror stuff fixed this afternoon
13:32:53mjn3are those the last issues holding up a bug release?
13:33:05anderseeyup
13:34:25mjn3good.  i'm ready to get back to the i10/i18 stuff
13:55:55mjn3andersee: btw, looks like the stange python failures with threads i was seeing last night were due to my fake librt.a  :-(  without it it works fine
13:59:17anderseecool.
13:59:25anderseeExcept for the fake librt that is...
14:00:54mjn3yeah.  downloading perl 5.8 now.  btw, have you looked at ccache.samba.org?  might be useful for testing on some of the slower platforms you have
14:08:08tzangermjn3: perl 5.8.0 defaults to using nm to find functions but then mistakenly finds isnan().  If you tell it to make tons of small programs to find the functions it works correctly
15:08:24mjn3tzanger: when you built perl 5.8, did you have uClibc built with thread support?
15:08:52tzangerno
15:09:00tzangeroh wait
15:09:04tzangerI thought you said perl
15:09:07tzangerhang on let me check :-)
15:09:52tzangeryes I have INCLUDE_THREADS = true in my uClibc Config file
15:09:53mjn3i ask because miniperl is segfaulting for me in build process...
15:09:55mjn3ok
15:09:57mjn3thanks
15:10:04tzangerI built perl without thread support though
15:10:32tzangeryeah miniperl was failing link for me when I let it use nm to find the supported library functions (the isnan() I told you about)
15:12:02mjn3different issue.  miniperl is linking fine.  however, i'm getting a segfault when the followin./miniperl -Ilib configpm configpm.tmpg is run in the build process:  
15:12:06mjn3./miniperl -Ilib configpm configpm.tmp
15:12:15tzangerhmm
15:12:19tzangerdid you compile perl with thread support?
15:12:23mjn3no
15:12:34tzangeroh
15:12:51mjn3i've built it with debugging turned on, and something is setting PL_strtab to NULL.  haven't tracked it down yet
15:13:05tzangerah I did not build with debugging
15:13:52mjn3i didn't the first time either... i did to try to isolate the segfault i saw when not debugging also
15:14:16tzangerhmm
15:14:21tzangerisn't software great?  :-)
15:21:58mjn3building with an older version of uClibc now to see if some of the recent changes might be the cause
15:23:28mjn3looks like... :-(
15:50:07tzangerif it helps, the snapshot from the 20th worked
16:52:25mjn3tzanger: i'm still getting a segfault with the snapshot from the 20th.  :-(  trying others
17:50:13tzangerhmm
17:50:18tzangerI definately am not...
17:50:24mjn3tzanger: did you build with large file support?
17:50:27tzangerno
17:51:13mjn3ok.  the problem i'm seeing is a result of a struct stat / struct stat64 mismatch when large file support is on
17:51:22tzangerah
17:51:37mjn3haven't figured out why though, since the uClibc headers seem to be doing the right thing
17:51:45tzangerI had fun with a few packages with lfs turned off... their "disable lfs" doesn't work
19:18:48mjn3andersee: i tracked down the perl problem.  struct stat and struct stat64 don't do the right thing when large file support is enabled.
19:19:19anderseeoh.
19:19:21andersee:(
19:19:35anderseeguess I screwed that up
19:20:17mjn3i'm trying to fix it now...
19:21:01anderseeI guess the #define I put in bits/stat.h wasn't working...
19:22:39mjn3it is a bit more complicated than that i'm afraid...
19:37:05mjn3hmm... i may have fixed it.. runing some tests
19:39:30anderseek
19:41:30mjn3seems to be working with large file support enabled now.  need to check without
19:41:39mjn3perl 5.8 has lots more tests
19:42:59mjn3good thing i didn't ever try this on that old 486 laptop...
19:43:40anderseehehe
19:43:40anderseeSpeed is a good thing sometimes. :)\
19:44:48anderseeSo did you change bits/kernel_stat.h to conditionally rename the structs based on __USE_LARGEFILE64?
19:45:50mjn3had to path perl 5.8 for a deprecated feature.  YESSTR and NOSTR are no longer supported, but perl tries to use them anyway it detects glibc
19:46:12mjn3ah... a sec
19:49:15mjn3i removed the #define stat stat64 stuff you added to common/bits/stat.h and i duplicated the struct stat64 declaration for struct statt if __USE_FILE_OFFSET64 was defined
19:49:54mjn3the python 2.2.1 tests passed.  i'll rebuild without large file support now
19:51:04mjn3hmm... before i do, pel 5.8 failed 5 tests.  4 involved the sysvipc stuff, but one involved exit... checking now
20:33:54mjn3andersee: i just checked in my mods to fix the struct stat/stat64 problem on i386.  take a look.  i haven't touched the other archs
21:42:06anderseeK
21:42:12anderseelooking now
21:42:18anderseetook a nap
21:48:26mjn3andersee: still haven't figures out what is failing in the perl 5.8 simple exit test...
21:48:41anderseeHmmm
21:49:15anderseeIt is testing anything special (atexit, etc)?  or just _exit()
21:49:47mjn3looking
22:07:32mjn3not sure what's going on... :-(
22:08:43mjn3both tests that failed exited with 0 when 255 was expected...
22:58:50anderseemjn3-away: ok, I got the shm stuff to work.
22:58:56anderseeCommitting to cvs now
23:00:15mjn3andersee: great.  i'll cvs up and check in bit
23:00:39anderseeTurned out to be not as painful as I'd expected...
23:01:17tzangerthat means that perl will install with uClibc without any failed tests
23:01:18tzangerwow
23:01:26anderseeI'd been procrastinating this one all week...
23:01:44anderseeI still need to fix bits/kernel_stat.h for all the other arches...
23:02:34anderseeok, committed the shm fix
23:16:25anderseeok, just committed the bits/kernel_stat.h cleanup
23:16:37anderseeSo in theory.... we are good to go
23:17:24mjn3andersee: i still need to fix the error message stuff.  i spent all day finding out why perl was segfaulting
23:17:33anderseeahh
23:17:46mjn3just started the perl tests
23:17:46anderseesorry about that...
23:19:03mjn3it was hard to track down because the variable in question was being overwritten by an fstat64 call since perl passed it a normal struct
23:19:15tzangermjn3: why was it?
23:19:25tzangerah
23:19:31tzangerI wrote too fast :-)
23:22:49anderseetesting now on powerpc...
23:27:58mjn3now 2 failures.  the simple exit test and one of the sysvipc msg tests
23:28:20anderseeWell, thats progress anyways
23:28:46mjn3yeah
23:29:23anderseeOdd that the exit test worked before...
23:29:59mjn3this is perl 5.8.  i was testing perl 5.6.1 before
23:30:05anderseeAhh
23:32:17mjn3failing msg test has to do with message queue number i think.
23:43:46mjn3andersee: i'll look into that single failure tomorrow i think.  i'm going to shut down for the day
23:47:23anderseemjn3: k. night then
23:47:29mjn3night

Generated by irclog2html.pl by Jeff Waugh - find it at freshmeat.net! Modified by Tim Riker to work with infobot logs, split per channel and by date, etc.