# 04:47:13 |
mnt_real |
quits : Quit: Leaving |
# 06:11:26 |
smartin |
joins #crosstool-ng |
# 09:11:54 |
smartin |
quits : Ping timeout: 252 seconds |
# 16:58:17 |
y_morin |
joins #crosstool-ng |
# 17:08:49 |
y_morin |
ccct, ccole: any one of you around? |
# 17:27:20 |
ccole |
y_morin: Hello! |
# 17:27:29 |
y_morin |
ccole: Heya! |
# 17:27:37 |
y_morin |
ccole: seen my messages yesterday? |
# 17:27:46 |
ccole |
y_morin: yes, sounds like good progress! |
# 17:28:06 |
ccole |
wonders why Ubuntu gang feels there's a need for yet another shell? |
# 17:28:27 |
y_morin |
ccole: because it is faster.. at breaking things! :-) |
# 17:30:34 |
ccole |
y_morin: I have found some interesting bugs in GCC and glibc which inhibit my using them. For example, rpc is broken in glibc2.14. And GCC4.6.0/1 produce incorrect asm code for ARM! |
# 17:31:04 |
ccole |
y_morin: These are known bugs.. but exist in the released tarballs at gnu.org. |
# 17:31:50 |
ccole |
y_morin: I finally found through steppnig back 1 version at a time that GCC 4.5.3 and glibc 2.13 are a stable combo for ARM. |
# 17:34:59 |
y_morin |
ccole: RPC is not broken. It;s just "removed" and deprecated in glibc. |
# 17:35:26 |
y_morin |
ccole: the (long-term) plan is to ramp up to speed on TI-RPC, but it's not ready for prime-time. |
# 17:36:02 |
y_morin |
ccole: did you try the Linaro gcc-4.6? IOt should be more stable on ARM. |
# 17:36:09 |
y_morin |
*It |
# 17:48:55 |
y_morin |
is now known as: y_morin|away |
# 17:52:56 |
y_morin|away |
is now known as: y_morin |
# 18:26:21 |
y_morin |
is now known as: y_morin|away |
# 19:28:14 |
y_morin|away |
is now known as: y_morin |
# 19:29:46 |
ccole |
y_morin: Thanks, I hadn't thought to look there; That appears to be a gold mine of patches for the ARM toolchain |
# 19:30:26 |
ccole |
y_morin: Basically linaro's own + stuff from CodeSourcery, which has not yet been put into GCC mainline |
# 19:31:07 |
y_morin |
ccole: yes, that's absically the workbench for ARM in gcc, before it gets pushed upstream. |
# 19:32:41 |
ccole |
y_morin: Awesome, thanks for enlightening me :) |
# 19:37:25 |
y_morin |
OK, so we are down to this situation, now: |
# 19:38:02 |
y_morin |
If '/bin/sh' points to 'dash', *and* CONFIG_SHELL is set (even to '/bin/sh' !), glibc build breaks |
# 19:38:50 |
y_morin |
If '/bin/sh' points to 'bash', *and* CONFIG_SHELL is empty, then it works. |
# 19:38:57 |
y_morin |
I have yet to test the following: |
# 19:39:26 |
y_morin |
1) Point '/bin/sh' to 'bash', and set CONFIG_SHELL to /bin/sh |
# 19:39:55 |
y_morin |
2) Point '/bin/sh' to 'bash', and set CONIG_SHELL to '/bin/bash' |
# 19:40:16 |
y_morin |
3) Point '/bin/sh' to 'dash', and set CONFIG_SHELL to '/bin/dash' |
# 20:15:26 |
y_morin |
is now known as: y_morin|away |
# 20:21:20 |
y_morin|away |
is now known as: y_morin |
# 20:34:26 |
smartin |
joins #crosstool-ng |
# 20:34:57 |
ccole |
y_morin: Interestingly, my Ubuntu 11.04 has sh -> dash, however it has SHELL=/bin/bash. That seems like a configuration mismatch to me.. |
# 20:36:00 |
y_morin |
ccole: grep 'your-login' /etc/passwd --> does it show /bin/bash, /bin/dash, or /bin/sh ? |
# 20:36:03 |
ccole |
y_morin: So perhaps I should try either your #2 or #3 and see if that will work. Both ought to I'd think. |
# 20:36:20 |
y_morin |
ccole: /bin/sh is the system shell, while it seems that your user-shell is bash. |
# 20:36:36 |
ccole |
y_morin: passwd says /bin/bash |
# 20:37:03 |
ccole |
y_morin: OK, I see. ps says Im running bash. |
# 20:37:06 |
y_morin |
ccole: well, I have another idea in mind: do as for autoconf, and run ./configure explicitly throught CONFIG_SHELL. |
# 20:38:08 |
y_morin |
*through |
# 20:50:13 |
y_morin |
OK, that last one is the simplest, and DOES ACTUALLY WORKS!!!! Yeah! :-) |
# 20:50:30 |
y_morin |
Expect soon a patch on your favorite website! :-) |
# 20:56:28 |
ccole |
y_morin: Great! Thank you! |
# 21:01:30 |
kos_tom |
y_morin: BTW, I successfully built and used a toolchain compiled statically with your python/gdb patch applied. But I haven't tested gdb/gdbserver yet. |
# 21:01:53 |
y_morin |
kos_tom: if it was statically linked, then python is not linked in. |
# 21:01:53 |
kos_tom |
but at least gcc/ld works, and is capable of building a kernel + a system with X.org and some other stuff, and the system runs :) |
# 21:02:07 |
y_morin |
kos_tom: That's good to read! :-) |
# 21:02:31 |
y_morin |
kos_tom: Note however, that bintuils were not entirely build static either. |
# 21:03:06 |
y_morin |
kos_tom: gold was always dynamically linked: http://crosstool-ng.org/hg/crosstool-ng/rev/293d8334b913 |
# 21:03:19 |
kos_tom |
ah, yes, the gold linker. |
# 21:04:01 |
kos_tom |
ok, I might add this patch as well, just to make sure, but I don't think I'll use the gold linker here. |
# 21:05:05 |
y_morin |
kos_tom: are you using the 1.12 series, or the devel stuff form the repo? |
# 21:05:36 |
kos_tom |
I'm using the 1.12.2 version plus the python/gdb patch. |
# 21:06:25 |
y_morin |
kos_tom: you may want to upgrade to 1.12.3, there are a few fixes: http://crosstool-ng.org/download/crosstool-ng/01-fixes/1.12.2/ |
# 21:06:59 |
y_morin |
(although the alternat kernel DL locations is now no longer working due to rsyncing an empty dir...) |
# 22:36:09 |
smartin |
quits |
# 22:56:23 |
y_morin |
quits : Quit: Night! |