| # 03:57:21 |
bhundven |
crazy patch madness |
| # 03:58:01 |
bhundven |
the patches on musl-cross don't all have the same support, version by version for gcc |
| # 03:59:32 |
bhundven |
finds himself diffing diffs. |
| # 05:49:28 |
feepbot |
quits : Remote host closed the connection |
| # 05:50:19 |
feepbot |
joins #crosstool-ng |
| # 06:00:04 |
feep_ |
is now known as: feep |
| # 06:06:37 |
diorcety |
joins #crosstool-ng |
| # 06:22:27 |
diorcety |
quits : Read error: Connection reset by peer |
| # 06:26:02 |
daggs1-work |
bhundven, any fix for the musl libc install bug? |
| # 06:27:17 |
bhundven |
nothing so far. I'm trying to fix up a new patch series |
| # 06:31:47 |
daggs1-work |
I've called the configure as said here before the headers install (had to move the arch setting to a function because the makefile complained arch isn't set) and it compile well but everything else still gets installed into sysroot/usr/{include,lib} |
| # 06:36:06 |
daggs1-work |
bhundven, here is what I used: http://codepad.org/X95RNwQD |
| # 06:36:25 |
daggs1-work |
ofcourse I removed the 0000.... patch |
| # 06:36:52 |
bhundven |
well |
| # 06:37:05 |
bhundven |
I keep getting distracted by life stuff |
| # 06:37:20 |
bhundven |
but last build after y_morin reworked threading |
| # 06:37:29 |
bhundven |
and I ran configure before install-headers |
| # 06:37:35 |
bhundven |
(to fix the path) |
| # 06:37:56 |
bhundven |
but crti.o and crtn.o are missing |
| # 06:38:08 |
bhundven |
think I'm forgetting something in my start files script |
| # 06:38:17 |
bhundven |
anyways |
| # 06:38:23 |
bhundven |
things are a bit different |
| # 06:38:34 |
bhundven |
so after I fix this stuff, try the new patches |
| # 06:40:39 |
daggs1-work |
I understand, ok includes seems to match the one in glibc |
| # 06:41:30 |
bhundven |
I can't diff it for some reason |
| # 06:41:37 |
bhundven |
I think it's the paste site you used |
| # 06:41:50 |
bhundven |
trailing spaces |
| # 06:42:01 |
daggs1-work |
but I'm concern about the libs install, it linked libc to /usr/lib or something |
| # 06:42:05 |
daggs1-work |
sec, will try another |
| # 06:42:44 |
daggs1-work |
this works? http://dpaste.com/38QMB24 |
| # 06:43:47 |
daggs1-work |
correction, ld-musl-x86_64.so.1 -> /usr/lib/libc.so |
| # 06:43:52 |
bhundven |
so I'm not really focused on the specific problem you're having, because I'm under the assumption that they are all broken with that last patch set. |
| # 06:44:11 |
daggs1-work |
no libc in sysroot/lib |
| # 06:44:18 |
bhundven |
I'm trying to make corrections to make the new patch set work, and then lets try again with the new patch set. |
| # 06:44:20 |
bhundven |
right |
| # 06:44:20 |
daggs1-work |
didn't said you should, just reportingh |
| # 06:44:34 |
bhundven |
Thank! I really appreciate it! |
| # 06:44:43 |
daggs1-work |
anything I can do to help speed it up? |
| # 06:45:00 |
bhundven |
well, everyone in my house is now sleeping |
| # 06:45:06 |
bhundven |
so no distractions :P |
| # 06:45:50 |
daggs1-work |
cannot say the same, my dogs keep bugging me and I need to go to work soon |
| # 06:45:52 |
bhundven |
less goalie mode, more code |
| # 06:47:32 |
bhundven |
hrm, why not leave do_get_arch global? |
| # 06:47:46 |
bhundven |
and just call it in the other spot |
| # 06:48:39 |
bhundven |
but then you don't call do_get_arch in do_libc |
| # 06:48:42 |
daggs1-work |
isn't that what I did? |
| # 06:48:51 |
bhundven |
so, I see your problem with musl trying to install to /usr/lib |
| # 06:49:03 |
bhundven |
instead of sysroot/lib |
| # 06:49:10 |
bhundven |
(right?) |
| # 06:49:21 |
bhundven |
there is no ARCH= |
| # 06:49:46 |
daggs1-work |
I see |
| # 06:50:09 |
daggs1-work |
so ARCH existing will fix the lib install? it seems that the include install is ok |
| # 06:50:14 |
bhundven |
I think you are seeing a different way that the same issue other toolchains are having (in their own way) |
| # 06:50:34 |
daggs1-work |
possibbly. |
| # 06:51:07 |
bhundven |
so let me keep focused on getting the updated patch set complete and posted to the ml, and try again. |
| # 06:51:33 |
daggs1-work |
ok |
| # 06:51:45 |
bhundven |
I didn't think of every possible outcome with the last patch set, and expected more critique with the RFC process |
| # 06:52:08 |
bhundven |
this time, I'm going to try a few builds before I post |
| # 06:52:19 |
daggs1-work |
ok |
| # 06:52:27 |
bhundven |
wishes he had a faster build box |
| # 06:52:47 |
daggs1-work |
:-P |
| # 06:53:57 |
bhundven |
also wishes sh4rm4 was here. He'd be happy to know I'm messing around with musl-cross |
| # 06:56:38 |
daggs1-work |
ok, I need to go, be back later |
| # 07:34:11 |
aalv |
joins #crosstool-ng |
| # 07:41:45 |
bhundven |
tlwoerner: running into a problem michael hope was trying to fix |
| # 07:42:48 |
bhundven |
https://sourceware.org/ml/crossgcc/2011-10/msg00047.html |
| # 07:43:33 |
bhundven |
need empty vendor string to test: arm-linux-musleabi instead of arm-unknown-linux-musleabi |
| # 07:45:14 |
bhundven |
never happened |
| # 07:45:15 |
bhundven |
:( |
| # 08:13:03 |
bhundven |
tlwoerner: his email address bounced and found he has moved on from linaro. So I guess that patch is in your hands. |
| # 08:13:05 |
bhundven |
? |
| # 08:36:18 |
voltagex1 |
kos_tom: you still around? |
| # 08:36:32 |
voltagex1 |
is now known as: voltagex |
| # 08:36:39 |
voltagex |
quits : Changing host |
| # 08:36:39 |
voltagex |
joins #crosstool-ng |
| # 08:36:41 |
kos_tom |
voltagex: yes. |
| # 08:36:42 |
daggs1-work |
bhundven, how can I instruct ct-ng to create a i386=>x86_64 when running the build on a x86_64 macvhine? |
| # 08:36:54 |
kos_tom |
voltagex: I'm actually based in Europe, so I'm generally around at this time :) |
| # 08:37:24 |
voltagex |
kos_tom: eglibc_2_8 contains a faulty check for make versions <=3.87, but I'm not sure how to write a patch for crosstool |
| # 08:37:38 |
voltagex |
kos_tom: I have enough trouble wrangling one time zone ;) |
| # 08:37:49 |
voltagex |
kos_tom: I'm UTC+10 btw |
| # 08:37:55 |
bhundven |
heh: [ALL ] /usr/src/tcwork/armtest/.build/arm-unknown-linux-musleabi/buildtools/lib/gcc/arm-unknown-linux-musleabi/4.9.0/../../../../arm-unknown-linux-musleabi/bin/ld: cannot open output file /usr/lib/libc.so: Permission denied |
| # 08:37:57 |
kos_tom |
voltagex: australia ? |
| # 08:38:14 |
bhundven |
woops |
| # 08:39:55 |
bhundven |
daggs1-work: BUILD/HOST/TARGET : you're treading in cross-native land |
| # 08:40:06 |
bhundven |
x86_64/i386/x86_64 |
| # 08:41:13 |
bhundven |
not really supported? |
| # 08:41:19 |
bhundven |
afaik |
| # 08:41:37 |
daggs1-work |
type should be cross native? |
| # 08:44:21 |
bhundven |
daggs1-work: I don't believe it is supported. |
| # 08:45:56 |
bhundven |
kos_tom: I have a question about musl support in buildroot |
| # 08:46:13 |
bhundven |
http://git.buildroot.net/buildroot/tree/package/musl/musl.mk#n63 |
| # 08:46:26 |
bhundven |
on line 63, you 'install-libs install-tools' |
| # 08:46:56 |
daggs1-work |
crap, that isn't good... |
| # 08:47:47 |
bhundven |
kos_tom: I guess, what is the difference between staging and target |
| # 08:48:04 |
bhundven |
line 63 vs. line 70 |
| # 08:48:12 |
kos_tom |
bhundven: target is what goes on the target |
| # 08:48:18 |
kos_tom |
bhundven: staging is all libraries/headers needed to build stuff |
| # 08:48:25 |
bhundven |
got it |
| # 08:48:27 |
kos_tom |
bhundven: staging is the toolchain sysroot |
| # 08:48:39 |
bhundven |
check. just wanted to make sure |
| # 08:50:32 |
bhundven |
I had it backwards :-/ hehe |
| # 08:50:54 |
bhundven |
almost 2am, mistakes like that are somewhat acceptable |
| # 09:01:19 |
bhundven |
so with the 2 step gcc build, don't have to do the crt[in].o stuff? |
| # 09:01:52 |
bhundven |
maybe that's next, but before I get back to multilib |
| # 09:06:00 |
bhundven |
daggs1-work: so you can image 4 types of compilers: native, cross, cross-native, and cross-canadian |
| # 09:06:26 |
bhundven |
native is build/host/target are all i386. crosstool-ng does not support this. |
| # 09:06:52 |
bhundven |
cross is crosstool-ng's most common target: x86_64/x86_64/arm |
| # 09:07:33 |
bhundven |
cross-native is not supported by crosstool-ng, but I'm sure someone could write support for it. x86_64/arm/x86_64 |
| # 09:08:11 |
daggs1-work |
bhundven, in my case, I need the i386 gcc libs and can use the -m32 switch |
| # 09:08:16 |
bhundven |
cross-canadian is somewhat supported in crosstool-ng, but needs some fixes to work right (iirc, there are some current bugs): x86_64/arm/powerpc |
| # 09:08:34 |
daggs1-work |
bhundven, not idea what cross-canadian ;) |
| # 09:08:38 |
bhundven |
idk why i choose those combinations, just examples |
| # 09:09:03 |
bhundven |
it builds on x86_64, runs on arm, builds target files for powerpc |
| # 09:09:10 |
bhundven |
(in that horrible example) |
| # 09:09:24 |
daggs1-work |
indeed... |
| # 09:09:37 |
bhundven |
idk why I'd build on arm for powerpc :p |
| # 09:09:46 |
daggs1-work |
will try to build on my lousy 2 cores desktop |
| # 09:14:53 |
bhundven |
daggs1-work: mine is atleast 4 cores, but i3 :-/ |
| # 09:15:10 |
daggs1-work |
mine is the beast! e8400! |
| # 09:15:13 |
bhundven |
i3-3110M |
| # 09:15:39 |
bhundven |
yea yea |
| # 09:15:41 |
bhundven |
:p |
| # 09:15:58 |
bhundven |
the system I had was an e5500 |
| # 09:16:10 |
bhundven |
bet the e8400 smokes |
| # 09:16:42 |
daggs1-work |
at one time I ran Gentoo on such system :) |
| # 09:16:55 |
bhundven |
gesh |
| # 09:17:26 |
bhundven |
pretty much have to have a system like that to run gentoo (unless you're doing binpkg) |
| # 09:17:45 |
bhundven |
building on another host, and doing binary install / stage3 |
| # 09:18:25 |
daggs1-work |
the e8400 was gentoo, the t8100 was binpkg :) |
| # 09:18:34 |
bhundven |
hah |
| # 09:18:38 |
daggs1-work |
now I run my gentoo on a i7-2600 16GB of ram |
| # 09:18:42 |
bhundven |
called it |
| # 09:19:43 |
bhundven |
I'm planning on getting a thinkserver rd440 and racking it somewhere, and dumping the i3 and my mac pro for a chromebook |
| # 09:27:10 |
daggs1-work |
well I build my server from parts, including my 6 hds (4 in raid) which serves as a multi seat box for 3 users (2 normal desktop and one htpc), just waiting for my airtame to arrive and I can finish the htpc part |
| # 09:29:02 |
bhundven |
meld confused me on your patch |
| # 09:29:09 |
bhundven |
you have it right |
| # 09:29:13 |
daggs1-work |
? |
| # 09:29:22 |
bhundven |
the one you posted earlier |
| # 09:29:41 |
bhundven |
A few more tests, and I'll email patches out |
| # 09:30:15 |
daggs1-work |
didn't understood the "have it right" part |
| # 09:30:58 |
daggs1-work |
gone for lunch, bbl |
| # 09:31:03 |
bhundven |
np |
| # 09:34:37 |
bhundven |
http://dpaste.com/38QMB24 |
| # 09:34:45 |
bhundven |
I read the diff backward |
| # 09:34:58 |
bhundven |
you broke out do_get_arch |
| # 09:35:14 |
bhundven |
I've integrated that change |
| # 09:50:38 |
bhundven |
[ERROR] /usr/src/tcwork/armtest/.build/src/gcc-4.9.0/libsanitizer/sanitizer_common/sanitizer_platform_limits_posix.cc:73:24: fatal error: net/if_ppp.h: No such file or directory |
| # 09:51:04 |
bhundven |
if_ppp.h is in sysroot/usr/include/linux/if_ppp.h |
| # 09:51:18 |
bhundven |
:/ |
| # 09:55:33 |
bhundven |
daggs1-work: I'm going to bed. my current wip is here: https://bitbucket.org/bhundven/crosstool-ng/branch/musl |
| # 09:55:53 |
bhundven |
complete with the above header issue |
| # 09:56:09 |
bhundven |
take a look and let me know. |
| # 11:03:30 |
sh4rm4 |
joins #crosstool-ng |
| # 11:07:05 |
sh4rm4 |
quits : Remote host closed the connection |
| # 11:08:49 |
sh4rm4 |
joins #crosstool-ng |
| # 11:11:24 |
sh4rm4 |
quits : Remote host closed the connection |
| # 11:17:01 |
sh4rm4 |
joins #crosstool-ng |
| # 11:22:46 |
sh4rm4 |
quits : Remote host closed the connection |
| # 11:23:16 |
sh4rm4 |
joins #crosstool-ng |
| # 11:40:36 |
sh4rm4 |
quits : Remote host closed the connection |
| # 11:42:48 |
daggs1-work |
bhundven, the only thing I can think of is that the compile cmd of that test doesn't includes sysroot/usr/include/linux/ |
| # 11:43:53 |
daggs1-work |
bhundven, but for glibc, it is in the same path so it might be wise to check how that works with glibc |
| # 11:45:45 |
sh4rm4 |
joins #crosstool-ng |
| # 12:06:50 |
maxime__ |
joins #crosstool-ng |
| # 12:09:00 |
maxime |
quits : Ping timeout: 255 seconds |
| # 12:13:22 |
maxime__ |
is now known as: maxime |
| # 12:13:25 |
maxime |
quits : Changing host |
| # 12:13:25 |
maxime |
joins #crosstool-ng |
| # 12:49:45 |
sh4rm4 |
quits : Remote host closed the connection |
| # 12:58:20 |
RushPL_ |
is now known as: RushPL |
| # 13:10:06 |
sh4rm4 |
joins #crosstool-ng |
| # 13:25:43 |
memfrob |
quits : Ping timeout: 272 seconds |
| # 13:34:17 |
sh4rm4 |
quits : Remote host closed the connection |
| # 13:36:08 |
sh4rm4 |
joins #crosstool-ng |
| # 13:37:59 |
memfrob |
joins #crosstool-ng |
| # 13:54:35 |
aalv1 |
joins #crosstool-ng |
| # 13:55:03 |
aalv |
quits : Read error: Connection reset by peer |
| # 13:58:19 |
daggs1-work |
bhundven, why are you creating libc.so and some o files in install header? doesn't that defies the install headers concept? |
| # 14:29:02 |
RushPL |
quits : Ping timeout: 250 seconds |
| # 14:29:38 |
RushPL |
joins #crosstool-ng |
| # 14:45:59 |
sh4rm4 |
quits : Ping timeout: 256 seconds |
| # 15:05:46 |
memfrob |
quits : Ping timeout: 260 seconds |
| # 15:07:03 |
loide |
joins #crosstool-ng |
| # 15:27:52 |
memleak |
joins #crosstool-ng |
| # 15:32:47 |
memleak |
quits : Ping timeout: 245 seconds |
| # 16:08:29 |
y_morin |
joins #crosstool-ng |
| # 16:34:47 |
loide |
quits : Ping timeout: 264 seconds |
| # 17:01:22 |
bhundven |
something about this musl setup is messing me up |
| # 17:08:40 |
nandub |
joins #crosstool-ng |
| # 17:10:51 |
bhundven |
I think I was running into a gcc-4.9.x bug or something |
| # 17:11:02 |
bhundven |
did the build with gcc-4.8.3, works |
| # 17:13:11 |
nandub |
quits : Client Quit |
| # 17:15:26 |
nandub |
joins #crosstool-ng |
| # 17:22:47 |
bhundven |
builds kernels. now time for some more platforms. |
| # 17:25:47 |
nandub |
quits : Quit: WeeChat 0.4.3 |
| # 17:26:08 |
loide |
joins #crosstool-ng |
| # 17:33:37 |
nandub |
joins #crosstool-ng |
| # 17:39:00 |
daggs1-work |
bhundven, question, did you encountered issues with busybox and musl? |
| # 17:39:26 |
bhundven |
haven't tried. let me check. |
| # 17:40:17 |
aalv1 |
quits : Quit: Leaving. |
| # 17:40:41 |
daggs1-work |
for some reason, when I try to run scripts, on that busybox I get exec format error |
| # 17:40:53 |
daggs1-work |
to be fair, I had is with uclibc too but not glibc |
| # 17:48:09 |
bhundven |
/usr/src/tcwork/busybox/scripts/gcc-version.sh: line 11: arm-eabi-gcc: command not found << this one? |
| # 17:49:51 |
bhundven |
is doing it wrong :P |
| # 18:00:05 |
bhundven |
I had a problem, but not an exec format error |
| # 18:00:13 |
bhundven |
I'm still having other issues. |
| # 18:00:56 |
daggs1-work |
bhundven, my issue is past ctng, I've built a cc, built busybox with it and I'm having issues with busybox |
| # 18:01:05 |
daggs1-work |
just wanted to know |
| # 18:01:34 |
bhundven |
I'm running into problems where the kernel headers are wrong. |
| # 18:01:43 |
bhundven |
I have to also go "fix someone's computer" :( |
| # 18:01:54 |
daggs1-work |
old same kernel as I use? |
| # 18:01:59 |
bhundven |
no |
| # 18:02:04 |
y_morin |
bhundven: Hello! Sorry, I was on the pthone, and now I'm off for dinner. I'll be back later tonight... |
| # 18:02:11 |
bhundven |
:sigh: |
| # 18:02:14 |
y_morin |
*phone... |
| # 18:03:19 |
bhundven |
I must just be dense or something. |
| # 18:03:22 |
bhundven |
https://bitbucket.org/bhundven/crosstool-ng/branch/musl |
| # 18:03:25 |
bhundven |
latest |
| # 18:04:19 |
bhundven |
I have to stop and go fix someone's computer where "stuff is missing" (windows xp? idk wtf i'm doing... wrt that. lol) |
| # 18:04:42 |
daggs1-work |
I feel your pain. |
| # 18:04:58 |
daggs1-work |
will take a look on the branch, any new changes? |
| # 18:05:14 |
bhundven |
daggs1-work, y_morin, anyone else... take a look at the branch |
| # 18:05:26 |
bhundven |
point out some obvious stupid problem I'm making. |
| # 18:05:36 |
bhundven |
my brain hurts |
| # 18:06:08 |
daggs1-work |
will try |
| # 18:06:53 |
diorcety |
joins #crosstool-ng |
| # 18:07:31 |
daggs1-work |
how do you clone that branch? |
| # 18:09:30 |
bhundven |
git clone https://bitbucket.org/bhundven/crosstool-ng.git |
| # 18:09:35 |
bhundven |
cd crosstool-ng |
| # 18:09:45 |
bhundven |
git checkout -b musl origin/musl |
| # 18:10:26 |
daggs1-work |
thanks |
| # 19:02:50 |
sh4rm4 |
joins #crosstool-ng |
| # 19:03:14 |
daggs1-work |
bhundven, you are missin the 0001-add_missing_signal_limits.patch in your branch |
| # 19:03:48 |
nandub |
quits : Quit: WeeChat 0.4.3 |
| # 19:10:18 |
bhundven |
sh4rm4: ok. this has been painful |
| # 19:10:47 |
sh4rm4 |
what's the issue ? |
| # 19:10:48 |
bhundven |
https://bitbucket.org/bhundven/crosstool-ng/branch/musl |
| # 19:11:17 |
bhundven |
I took the patches from musl-libc, instead of the patches that timo provided |
| # 19:11:50 |
bhundven |
had problems with missing crt/crt*.o files, but that is because we use the three step gcc build. |
| # 19:12:06 |
bhundven |
I'm hoping after getting musl in, we can start work on two step build |
| # 19:12:28 |
bhundven |
I know some work has been done to start 2 step |
| # 19:12:57 |
bhundven |
now I ran into something random |
| # 19:13:21 |
bhundven |
headers installing wrong |
| # 19:13:24 |
sh4rm4 |
what's the problem with the kernel header location mentioned in the commit message ? |
| # 19:13:53 |
bhundven |
gcc freaks out on not being able to include |
| # 19:14:05 |
bhundven |
find it in the toolchain under: linux/if_ppp.h |
| # 19:14:20 |
bhundven |
instead of linux/net/if_ppp.h |
| # 19:14:31 |
sh4rm4 |
btw it's recommended to use the patched kernel-headers pkg from sabotage linux (also used by musl-cross) |
| # 19:15:10 |
sh4rm4 |
because the kernel headers are incompatible with userspace in subtle ways; but for GLIBC they have an (ugly) workaround |
| # 19:15:10 |
bhundven |
man... |
| # 19:15:16 |
sh4rm4 |
(libc-compat.h) |
| # 19:15:36 |
sh4rm4 |
ah |
| # 19:15:49 |
sh4rm4 |
that's probably also the reason musl doesnt need a 3-stage build |
| # 19:16:00 |
sh4rm4 |
it doesnt depend on external kernel headers |
| # 19:16:17 |
sh4rm4 |
while glibc just #include_next the kernel headers |
| # 19:16:32 |
sh4rm4 |
gcc freaks out on not being able to include |
| # 19:16:39 |
sh4rm4 |
never heard about this issue |
| # 19:16:49 |
bhundven |
yea. I think I messed something up |
| # 19:16:58 |
bhundven |
nothing is poping out at me |
| # 19:18:12 |
bhundven |
everyone said multilib would be harder to implement in crosstool-ng... I'm finding musl more painful. |
| # 19:18:19 |
daggs1-work |
bhundven, linux headers installed ok here using kernel 3.12.5 |
| # 19:20:13 |
sh4rm4 |
https://github.com/GregorR/musl-cross/pull/35/files |
| # 19:20:15 |
bhundven |
gcc, the kernel, and many other tools don't support it without patches. not that musl should conform, but getting support for it in upstream tools would make things *WAY* easier. |
| # 19:20:28 |
sh4rm4 |
https://github.com/sabotage-linux/kernel-headers |
| # 19:21:19 |
sh4rm4 |
ah, so you mean we should ship broken stuff like glibc that depends on complicated workarounds, so just that you can keep doing workarounds as used to ? :) |
| # 19:21:29 |
bhundven |
haha, no |
| # 19:21:55 |
sh4rm4 |
you dont really need to use the patched kernel headers |
| # 19:22:33 |
sh4rm4 |
however you will probably run into some redefinitions of struct ethhdr and similar things when trying to build system software like busybox, iptables, etc |
| # 19:22:44 |
bhundven |
yes. I ran into that |
| # 19:23:03 |
sh4rm4 |
that's mostly fixed in the kernel-headers package i linked to |
| # 19:23:35 |
sh4rm4 |
one caveat exists still tho: in order for libc-compat.h to kick in, *all* kernel headers need to be included after the userspace headers |
| # 19:23:58 |
sh4rm4 |
but 90% of things do so and just work |
| # 19:24:06 |
sh4rm4 |
the other 10% need some rejuggling |
| # 19:24:18 |
bhundven |
the problem is... as currently being discussed in #musl |
| # 19:24:21 |
bhundven |
"12:08:13 <@dalias> if you have a patch whose context is purely boilerplate or other trivial content, and where all the + lines are completely new content, you would have a really hard time arguing that the patch is a derivative work" |
| # 19:25:02 |
bhundven |
and keeping patches in crosstool-ng adds non-technical problems |
| # 19:25:36 |
bhundven |
it would be best if the changes needed by those patches and custom kernel headers was upstream |
| # 19:25:45 |
sh4rm4 |
haha |
| # 19:25:59 |
sh4rm4 |
the patches were sent and ignored |
| # 19:26:10 |
bhundven |
because you didn't follow gcc's process |
| # 19:26:18 |
bhundven |
not *you* but who sent them |
| # 19:26:22 |
sh4rm4 |
talking about kernel-headers now |
| # 19:27:08 |
bhundven |
regardless. if kernel headers should be fixed in linux to handle properly for musl, why not push to get the changes upstream? |
| # 19:27:27 |
sh4rm4 |
because the maintainers of libc-compat.h are redhat guys |
| # 19:27:46 |
sh4rm4 |
that's why any request regarding musl is ignored |
| # 19:28:04 |
bhundven |
well, then the ultimate question is: Is there a business case for musl-libc, outside of redhat's interests? |
| # 19:28:10 |
bhundven |
My answer is yes |
| # 19:28:32 |
bhundven |
if so, bump up the chain of command |
| # 19:28:47 |
bhundven |
make the business case |
| # 19:29:19 |
bhundven |
imo, redhat doesn't run gnu/linux. the fsf/gnu do |
| # 19:29:38 |
sh4rm4 |
well i tried it, and it was ignored. i dont have the nerves to keep pushing and pushing |
| # 19:30:20 |
sh4rm4 |
(i also sent patches to busybox, which fix the above mentioned kernel header inclusion order, was ignored as well) |
| # 19:39:49 |
daggs1-work |
bhundven, have you encountered an issue with _ISdigit being missing? |
| # 19:41:25 |
bhundven |
daggs1-work: all sorts of problems right now |
| # 19:41:32 |
bhundven |
so, I'm sure you are too :) |
| # 19:42:19 |
daggs1-work |
strange, my wife keeps telling me that... |
| # 20:00:19 |
y_morin |
bhundven: Ping? |
| # 20:00:27 |
bhundven |
y_morin: pong |
| # 20:00:45 |
y_morin |
bhundven: https://bitbucket.org/bhundven/crosstool-ng/commits/5d5518851d0f54d78b671859d9153e1b2d06c710?at=master#Lscripts/build/libc/musl.shT57 <- should that not be: --host=${CT_BUILD} instead? |
| # 20:01:16 |
bhundven |
an experiment I forgot to remove |
| # 20:01:24 |
y_morin |
bhundven: OK. |
| # 20:01:39 |
bhundven |
interesting to note. that is how buildroot does it. |
| # 20:01:52 |
bhundven |
host and target both are the target |
| # 20:02:08 |
y_morin |
Ah... |
| # 20:02:52 |
y_morin |
Right... |
| # 20:03:17 |
bhundven |
I'm in #musl trying to sort out the patch stuff. And trying to see if I can help at all. |
| # 20:03:18 |
daggs1-work |
bhundven, got it, the macros are missing |
| # 20:03:34 |
daggs1-work |
from ctype.h |
| # 20:03:51 |
bhundven |
daggs1-work: my patch is wrong |
| # 20:03:57 |
daggs1-work |
which? |
| # 20:04:08 |
bhundven |
I'm not sure, but it's broken |
| # 20:04:34 |
y_morin |
bhundven: If you do not need the two gcc core passes, you can play with either of: CT_CC_CORE_PASS_{1,2}_NEEDED |
| # 20:04:52 |
bhundven |
yea |
| # 20:05:31 |
y_morin |
bhundven: And then carefully tweak the gcc script to do what you need in the needed pass. |
| # 20:06:20 |
y_morin |
bhundven: Especially wrt to (threading-model)=>(static-or-shared,build-libgcc-or-not) |
| # 20:07:25 |
y_morin |
bhundven: (but maybe CT_THREAD should be 'musl' in that case.) |
| # 20:17:22 |
bhundven |
I'm going to be away for a bit, but lurking. |
| # 20:19:41 |
y_morin |
bhundven: K. |
| # 20:25:36 |
daggs1-work |
going to bed, see you all in a few hours |
| # 20:25:46 |
y_morin |
daggs1-work: Night! |
| # 20:26:28 |
daggs1-work |
night |
| # 20:36:33 |
sh[4]rm4 |
joins #crosstool-ng |
| # 20:37:19 |
sh4rm4 |
quits : Ping timeout: 256 seconds |
| # 20:40:34 |
sh[4]rm4 |
is now known as: sh4rm4 |
| # 21:05:12 |
y_morin |
bhundven: If you're still lurking around ;-) care you share your defconfig for a musl toolchain? |
| # 21:05:23 |
loide |
quits : Ping timeout: 264 seconds |
| # 21:19:43 |
y_morin |
Gah... Building on this slow laptop is a pain... :-( |
| # 21:40:18 |
y_morin |
One very good point to musl: it compiles without a single warning! :-) |
| # 21:41:50 |
diorcety |
quits : Quit: Leaving. |
| # 21:47:18 |
y_morin |
Ah... But no rpc.h ... :-/ |
| # 21:48:03 |
y_morin |
Not that I care much about, anyway... |
| # 21:57:37 |
sh4rm4 |
quits : Remote host closed the connection |
| # 22:00:36 |
sh4rm4 |
joins #crosstool-ng |
| # 22:09:13 |
sh4rm4 |
quits : Remote host closed the connection |
| # 22:12:04 |
sh4rm4 |
joins #crosstool-ng |
| # 22:17:00 |
y_morin |
bhundven: You may want to include that in your tree: http://code.bulix.org/cji9rr-86597?raw |
| # 22:18:09 |
y_morin |
bhundven: With that, I have a (semmingly?) working toolchain that can build a large part of busybox. |
| # 22:18:50 |
y_morin |
bhundven: It still fails on ifplugd because of redefinition of struct ethhdr, so that must be fixed either in musl or the kernel headers. |
| # 22:19:22 |
y_morin |
bhundven: Also, musl lacks rpc/rpc.h, but that can be provided by libti-rpc, so is not a blocking factor to me. |
| # 22:19:52 |
y_morin |
bhundven: So, I guess we're approaching a state where we could apply it. :-) |
| # 22:26:16 |
mingwandroid |
quits : Remote host closed the connection |
| # 22:32:23 |
y_morin |
quits : Quit: Nighty Night! |
| # 23:21:25 |
maxime__ |
joins #crosstool-ng |
| # 23:23:14 |
maxime |
quits : Ping timeout: 250 seconds |