# 01:07:50 |
kpet |
quits : Ping timeout: 252 seconds |
# 03:23:39 |
mnt_real |
quits : Quit: Leaving |
# 09:56:23 |
y_morin |
joins #crosstool-ng |
# 10:35:32 |
esbenh |
what is the opinion on adding a patch for re-enabling RPC support in glibc 2.14? |
# 10:36:15 |
esbenh |
as I understand it, ti-rpc simply is not ready as a viable replacement yet, leaving glibc 2.14+ practically useless in a larger context |
# 10:37:12 |
esbenh |
I took a patch like that from eglibc 2.14 recipe of OE |
# 10:37:46 |
esbenh |
works nicely with glibc 2.14 with reformatting |
# 10:48:53 |
y_morin |
is now known as: y_morin|away |
# 12:05:38 |
smartin |
joins #crosstool-ng |
# 12:22:27 |
y_morin|away |
is now known as: y_morin |
# 12:23:28 |
y_morin |
esbenh: I have no strong opinion for or against. |
# 12:24:20 |
y_morin |
esbenh: Of course, this would be a workaround, not a fix. |
# 12:25:01 |
y_morin |
esbenh: The proper fix would be to either beat the glibc folks to revise their position, or to get ti-rpc up-to-speed. |
# 12:47:48 |
esbenh |
hehe, I agree. But... |
# 12:48:43 |
esbenh |
I don't have time or motivation to help ti-rpc get up to speed, and I don't envision anyone being able to change the position of the glibc folks, at least not mr. drepper |
# 13:03:31 |
y_morin |
is now known as: y_morin|away |
# 13:06:16 |
kos_tom |
it is kind of amazying to remove the RPC support from glibc if ti-rpc is not ready |
# 13:49:19 |
smartin |
quits |
# 14:46:00 |
esbenh |
even it it was ready, it is a different api, so completely dropping sunrpc is a rather drastic approach. |
# 14:58:09 |
y_morin|away |
is now known as: y_morin |
# 14:58:37 |
y_morin |
kos_tom: heya! |
# 14:58:50 |
y_morin |
kos_tom: I've looked at linking statically against libpython. |
# 14:59:09 |
y_morin |
kos_tom: seems there are abunch of issues doing so. |
# 14:59:34 |
kos_tom |
ah ah ? |
# 15:00:14 |
y_morin |
kos_tom: lemme dig up the pointers I found yesterday... |
# 15:00:41 |
y_morin |
kos_tom: 1st one: http://comments.gmane.org/gmane.comp.gdb.devel/30708 |
# 15:02:08 |
y_morin |
kos_tom: 2nd one: http://sourceware.org/ml/gdb/2010-01/msg00223.html |
# 15:02:29 |
y_morin |
that 2nd one says "Static linking may be a problem." |
# 15:03:52 |
y_morin |
kos_tom: probably we could work it around, or probably later versions behave better, but it does not look as easy as just adding -static to the linker flags |
# 15:04:09 |
kos_tom |
another option would be to build dynamically against python (and maybe gmp, mpfr, etc.) and build the gcc, gdb and al. binaries with an rpath |
# 15:04:37 |
y_morin |
kos_tom: don't talk to me about rpath. I don't want to hear about it. rpath is evil. |
# 15:04:42 |
y_morin |
:-) |
# 15:04:43 |
kos_tom |
if you add -static to the link flags, it builds it completely statically (even against glibc). I think linking statically against libpython is enough. |
# 15:04:48 |
kos_tom |
y_morin: I don't understand why. |
# 15:05:00 |
kos_tom |
for what ct-ng is doing, rpath is really a good option. |
# 15:05:16 |
y_morin |
kos_tom: if you use rpath, you can't relocate. |
# 15:05:20 |
kos_tom |
yes you can |
# 15:05:30 |
kos_tom |
ever heard of $ORIGIN ? |
# 15:05:48 |
y_morin |
kos_tom: Ah, yes. $ORIGIN. |
# 15:06:17 |
kos_tom |
with $ORIGIN, you can relocate. The libs just need to be at some fixed location relative to the binaries, but this isn't a problem. |
# 15:07:09 |
kos_tom |
the fact that gdb isn't linked statically against libpython makes it practically completely useless to link gcc statically against mpfr/ppl/gmp, because it means that you can't move the toolchain to a different host. |
# 15:07:54 |
y_morin |
kos_tom: the static gmp/mpfr/ppl/... is not originally about moving across hosts, but to ease relocatability. |
# 15:08:06 |
y_morin |
kos_tom: it is to avoid needing LD_LIBRARY_PATH. |
# 15:08:48 |
y_morin |
kos_tom: but don't get me wrong! I do agree that crossing host-boundary can be useful. |
# 15:08:59 |
y_morin |
*usefull. |
# 15:10:01 |
y_morin |
kos_tom: there's nothing about @ORIGIN in my ld man page. There is a reference to 'origin' in the -z option, but not in the -rpath opiton. |
# 15:10:34 |
kos_tom |
man ld.so |
# 15:10:48 |
kos_tom |
$ORIGIN |
# 15:10:48 |
kos_tom |
ld.so understands the string $ORIGIN (or equivaâ |
# 15:10:48 |
kos_tom |
lently ${ORIGIN}) in an rpath specification to |
# 15:10:48 |
kos_tom |
mean the directory containing the application |
# 15:10:49 |
kos_tom |
executable. |
# 15:10:55 |
y_morin |
Ah, good. |
# 15:11:19 |
kos_tom |
but as you don't build libpython, you would have to copy the host libpython.so there. |
# 15:11:46 |
kos_tom |
another option is to try to link statically against libpython.a, see what the problems are and see if it's doable to fix them. |
# 15:11:50 |
y_morin |
kos_tom: indeed. And I do *not* want to build libpython. :-) |
# 15:12:07 |
y_morin |
kos_tom: indeed. That's what I've started... |
# 15:12:20 |
kos_tom |
great! |
# 15:13:04 |
y_morin |
kos_tom: I was just saying there's a not-completely-paved-yet road ahead... |
# 15:14:27 |
y_morin |
kos_tom: and embedding the host libpython.so is not really an option either. It has dependencies: openssl, libz. |
# 15:15:52 |
y_morin |
kos_tom: quote from a stack-overflow post: "Whoever decided to make the token $ORIGIN is an evil bastard who deserves a special place in programmer hell." |
# 15:15:56 |
y_morin |
Lol! :-) |
# 15:19:16 |
kos_tom |
:) |
# 15:19:49 |
y_morin |
kos_tom: do you know if the $ORIGIN trick works with uClibc's ld.so ? |
# 15:37:57 |
kos_tom |
I am not sure. |
# 15:39:56 |
y_morin |
kos_tom: looks like not. I've looked at the code, and there are references to it, but no actual use of the #defined values... |
# 15:40:33 |
kos_tom |
yeah, same conclusion her. |
# 15:40:52 |
kos_tom |
but do you expect people to run ct-ng generated toolchains on uclibc based hosts ? |
# 15:41:49 |
y_morin |
kos_tom: Yes, I do. I'm planning to do so in a short time. If I can get enough free time... |
# 15:44:57 |
y_morin |
But that's not a show-stopper for now! |
# 15:45:22 |
kos_tom |
ok, I'm leaving. |
# 15:45:50 |
y_morin |
See ya! |
# 16:41:36 |
smartin |
joins #crosstool-ng |
# 17:09:48 |
y_morin |
is now known as: y_morin|away |
# 18:46:02 |
smartin |
quits |
# 19:58:18 |
y_morin|away |
is now known as: y_morin |
# 20:14:40 |
kos_tom |
y_morin: have you found any solution? |
# 20:15:16 |
y_morin |
kos_tom: not really worked on it since last time we spoke... I just came back after dinner. |
# 20:16:14 |
kos_tom |
ok. |
# 20:16:15 |
y_morin |
kos_tom: and I have a bunch of pending patches I have to handle first ( contributors deserve a bit of my time, don't they? :-) ) |
# 20:18:01 |
y_morin |
kos_tom: Something that just sprung to my mind: there's already the "static toolchain" config option, which builds a totally staticaly linked toolchain. |
# 20:18:20 |
y_morin |
kos_tom: Maybe that's what should be used by people that want to move a toolchain across hosts... |
# 20:19:26 |
y_morin |
kos_tom: My reasonning here, is that if one needs to move the toolchain, one knows that fact, and acts accordingly; that is, set "static toolchain". |
# 20:35:39 |
kos_tom |
y_morin: but does it build statically against libpython ? |
# 20:35:44 |
kos_tom |
(and does it work) ? |
# 20:36:04 |
kos_tom |
but IMO, it'd be great to be able to build dynamically against the C library and statically against everything else. |
# 20:36:19 |
kos_tom |
that's how CodeSourcery toolchain work, as far as I can understand. |
# 20:42:33 |
y_morin |
kos_tom: http://code.bulix.org/a8x6uv-80667 |
# 20:42:42 |
y_morin |
kos_tom: python is not supported in CS toolchains. |
# 20:48:23 |
kos_tom |
really ? I had the feeling that in recent versions of gdb, python was mandatory to do some remote debugging, because the target descriptions are written in XML and parsed using Python. |
# 20:48:50 |
kos_tom |
at least, in our training sessions, when gdb is built without python support, remote debugging support doesn't work ("remote g packet too long") |
# 20:49:12 |
y_morin |
kos_tom: well, that's what it says in the snippet I pasted above... |
# 20:49:23 |
kos_tom |
yeah, I see. |
# 20:49:53 |
y_morin |
kos_tom: I have to admit it's been quite some time I needed remote debugging... |
# 20:49:57 |
kos_tom |
aah noo, sorry. |
# 20:50:06 |
kos_tom |
it's when expat is not detected that there is a problem. |
# 20:50:12 |
kos_tom |
which looks more logical. |
# 20:50:15 |
y_morin |
kos_tom: Ah, ah! |
# 20:50:26 |
kos_tom |
see my email on the crossgcc list Mon, 25 Jul 2011 12:07:09 +0200 |
# 20:50:28 |
y_morin |
kos_tom: IIRC, python is for pretty-printing stuff... |
# 20:50:43 |
kos_tom |
y_morin: would it be possible to have a ct-ng option to build without python support ? |
# 20:50:56 |
y_morin |
kos_tom: definitely, yes. |
# 20:50:56 |
kos_tom |
(I guess you'll say: send a patch :-) ) |
# 20:51:29 |
y_morin |
kos_tom: actually, I was going to say: well, hm... Err... Lemme see... |
# 20:51:31 |
y_morin |
;-) |
# 20:53:14 |
kos_tom |
--with-python=no ? |
# 20:53:28 |
y_morin |
kos_tom: yes, seems like it. |
# 20:54:02 |
y_morin |
kos_tom: it can also be used to point to an alternate python (executable or install prefix) |
# 20:54:50 |
kos_tom |
hm, gdb also depends on expat, so it should also be statically linked against expat |
# 20:55:40 |
y_morin |
kos_tom: well, mine also links to openssl (libssl.so) |
# 20:56:00 |
y_morin |
kos_tom: and ncurses, and libz. |
# 20:56:45 |
y_morin |
kos_tom: http://code.bulix.org/1ob4hg-80668 |
# 21:04:29 |
y_morin |
kos_tom: what about: http://code.bulix.org/7bf8t7-80669 |
# 21:09:00 |
kos_tom |
yeah, sounds good. |
# 21:09:49 |
kos_tom |
the cs gdb is dynamically linked against ncurses, but seems to be linked statically against expat and all other libraries. |
# 21:16:59 |
ccct |
y_morin: Any new insight on why compiling glib-ports loops on ubuntu? (problem from 2 days ago) |
# 21:18:02 |
y_morin |
is now known as: y_morin|away |
# 21:18:31 |
y_morin|away |
is now known as: y_morin |
# 21:18:44 |
y_morin |
ccct: no, not yet... |
# 21:19:10 |
y_morin |
kos_tom: indeed. |
# 21:32:41 |
y_morin |
ccct: what version of ubuntu was it failing in ? |
# 21:33:07 |
y_morin |
ccct: can you pastebin your /etc/apt/sources.list, please? |
# 21:34:18 |
ccct |
y_morin: old intrepid |
# 21:34:39 |
y_morin |
ccct: i386 or amd64 ? |
# 21:35:20 |
ccct |
i386 |
# 21:36:42 |
y_morin |
Arg. Intrepid is no longer avail on archive.ubuntu.com Sigh... |
# 21:38:14 |
y_morin |
ccct: can you pastebin your /etc/apt/sources.list please? |
# 21:38:28 |
ccct |
http://crosstool-ng.org/download/ibot-logs/2011-07-03.html |
# 21:38:35 |
ccct |
That guy was using kubuntu 11.04 |
# 21:38:46 |
ccct |
http://pastebin.com/HNMGkUmp < sources.list |
# 21:39:12 |
y_morin |
ccct: Ah, old-releases, it is... |
# 21:40:00 |
y_morin |
OK, debootstrap-ing in progress... |
# 22:17:16 |
y_morin |
kos_tom: how did you manage your 32-bit chroot on your 64-bit machine, in the end? |
# 22:56:59 |
y_morin |
quits : Quit: Night... Sleep... |
# 23:48:21 |
mnt_real |
joins #crosstool-ng |