| 00:00.17 | *** join/#nslu2-linux z0mbi3 (~z0mbi3@a81-84-140-50.netcabo.pt) |
| 00:00.37 | jacques | hmm io.c first includes awk.h which includes config.h which defines HAVE_SYS_PARAM_H=1, and then io.c is supposedly including sys/param.h |
| 00:01.53 | jacques | but param.h would have to be included before awk.h in order for NGROUPS_MAX to be defined when awk.h decides whether to declare ngroups and groupset |
| 00:05.44 | jacques | trying something hackish |
| 00:06.12 | jacques | first I wanna get it compiled and see if it works better than the older version, then I'll worry about the include file issues |
| 00:06.39 | kergoth | add the new awk to oe and send me a patch while you're at it |
| 00:06.50 | kergoth | bleh, this headache just wont go away |
| 00:07.28 | jacques | ok will do (if it works any better :-) |
| 00:16.08 | jacques | hmm marginally better on gawk's own tests - fails 3 out of about 25 and then hangs on 'clobber' test |
| 00:16.24 | jacques | same gawk on my PC passes all tests |
| 00:20.39 | jacques | still failing a ton of tests (I removed clobber test to see how much further it would get) and then it eventually runs out of memory (this with 64MB swap) on the funstack test and OOM killer gets it |
| 00:21.56 | jacques | sigh it still fails on the perl configure version extraction :-( |
| 00:22.57 | jacques | ah well I'm out of ideas because everything else is too scary to think about - like some problem with glibc |
| 00:23.12 | jacques | (which would explain why it works with uclibc) |
| 00:23.30 | [g2] | jacques, can you increase you swap space as a test ? |
| 00:23.32 | jacques | but there are so many variables that I can only speculate |
| 00:23.42 | jacques | [g2], that won't accomplish anything |
| 00:23.48 | jacques | useful IMHO |
| 00:24.10 | [g2] | because you think it'll fail the same way |
| 00:24.17 | jacques | it should not be taking that much RAM - it's a runaway, and increasing swap would only make it take longer to fail |
| 00:25.04 | jacques | and the OOM thing is the least of my worries - it's failing tons of tests and the perl configure test without running out of RAM |
| 00:25.08 | [g2] | is there a way to run it on the PC and see how much memory it utilizes ? |
| 00:25.19 | jacques | please forget I said anything about the OOM |
| 00:25.24 | jacques | total red herring |
| 00:25.29 | [g2] | nod. |
| 00:25.37 | [g2] | back to 3.1.2 ? |
| 00:25.46 | [g2] | or this was 3.1.2 |
| 00:26.27 | jacques | this was 3.1.4, but I think it's something else causing it to fail like the libc - 3.1.2 worked in uclibc buildroot-built devel env |
| 00:26.44 | jacques | I don;t see why 3.1.4 would have regressed that much |
| 00:27.00 | jacques | so I conclude it's something other than the version of gawk which is causing the problems |
| 00:28.02 | jacques | I'm a little suspicious that busybox awk fails in the exact same way on the perl configure test I am doing |
| 00:28.14 | jacques | the test is: |
| 00:28.18 | jacques | awk '/define[ ]+PERL_SUBVERSION/ {print $3}' patchlevel.h |
| 00:28.28 | jacques | in the perl source dir which has patchlevel.h |
| 00:28.48 | jacques | it's supposed to return a number like 5 |