# 00:10:44 |
bhundven |
tlwoerner: ? |
# 01:44:35 |
djerome |
joins #crosstool-ng |
# 01:57:51 |
diorcety |
joins #crosstool-ng |
# 02:00:24 |
diorcety1 |
quits : Ping timeout: 240 seconds |
# 02:11:18 |
tlwoerner |
bhundven: :-) |
# 03:30:19 |
perr |
joins #crosstool-ng |
# 03:30:19 |
perr |
quits : Changing host |
# 03:30:19 |
perr |
joins #crosstool-ng |
# 03:36:50 |
perr |
quits : Ping timeout: 264 seconds |
# 03:46:40 |
perr |
joins #crosstool-ng |
# 03:53:37 |
perr |
quits : Ping timeout: 265 seconds |
# 04:14:44 |
perr |
joins #crosstool-ng |
# 04:19:21 |
bhundven |
gah, I've been sick for 3 days. I keep missing people on chat. |
# 04:29:38 |
perr |
quits : Ping timeout: 240 seconds |
# 04:42:03 |
perr |
joins #crosstool-ng |
# 05:15:06 |
djerome |
quits : Remote host closed the connection |
# 05:28:52 |
tlwoerner |
bhundven: still here? |
# 05:30:38 |
perr |
quits : Ping timeout: 240 seconds |
# 05:32:17 |
tlwoerner |
email probably works better :-) |
# 05:43:39 |
perr |
joins #crosstool-ng |
# 06:20:38 |
perr |
quits : Ping timeout: 240 seconds |
# 06:34:09 |
perr |
joins #crosstool-ng |
# 07:51:15 |
diorcety |
quits : Quit: Leaving. |
# 07:55:36 |
perr |
quits : Ping timeout: 240 seconds |
# 08:02:15 |
hurdman |
tu avais quand même l'usb2serial ? |
# 08:02:22 |
hurdman |
sorry bad chan |
# 08:11:59 |
perr |
joins #crosstool-ng |
# 08:12:17 |
hurdman |
is now known as: nantes_geek |
# 08:12:50 |
nantes_geek |
is now known as: hurdman |
# 08:19:26 |
perr |
quits : Ping timeout: 264 seconds |
# 08:43:20 |
aalv |
joins #crosstool-ng |
# 08:48:17 |
perr |
joins #crosstool-ng |
# 08:56:40 |
perr |
quits : Ping timeout: 265 seconds |
# 09:27:08 |
silviof |
joins #crosstool-ng |
# 09:27:08 |
perr |
joins #crosstool-ng |
# 09:27:09 |
perr |
quits : Changing host |
# 09:27:09 |
perr |
joins #crosstool-ng |
# 09:27:21 |
silviof |
parts #crosstool-ng |
# 09:48:17 |
duffduff |
parts #crosstool-ng |
# 10:33:46 |
Net147 |
joins #crosstool-ng |
# 11:10:00 |
perr |
quits : Ping timeout: 240 seconds |
# 11:22:09 |
perr |
joins #crosstool-ng |
# 11:40:52 |
zkrx |
joins #crosstool-ng |
# 11:54:48 |
zkrx |
quits : Ping timeout: 240 seconds |
# 11:59:58 |
zkrx |
joins #crosstool-ng |
# 12:48:58 |
zkrx |
quits : Ping timeout: 240 seconds |
# 12:54:06 |
zkrx |
joins #crosstool-ng |
# 13:07:18 |
perr |
quits : Ping timeout: 240 seconds |
# 14:02:14 |
perr |
joins #crosstool-ng |
# 14:42:27 |
Net147 |
quits : Quit: HydraIRC -> http://www.hydrairc.com <- |
# 15:00:48 |
zkrx |
quits : Remote host closed the connection |
# 15:02:54 |
zkrx |
joins #crosstool-ng |
# 15:21:46 |
mnt_real |
quits : *.net *.split |
# 15:29:11 |
mnt_real |
joins #crosstool-ng |
# 16:12:14 |
perr |
quits : Ping timeout: 264 seconds |
# 16:40:13 |
mingwandroid |
joins #crosstool-ng |
# 16:50:29 |
mingwandroid |
quits : Read error: Connection reset by peer |
# 16:51:41 |
mingwandroid |
joins #crosstool-ng |
# 16:54:47 |
zkrx |
quits : Ping timeout: 272 seconds |
# 17:17:06 |
y_morin |
joins #crosstool-ng |
# 17:22:45 |
zkrx |
joins #crosstool-ng |
# 17:26:37 |
mingwandroid |
quits : Read error: Connection reset by peer |
# 17:27:51 |
mingwandroid |
joins #crosstool-ng |
# 17:33:03 |
aalv |
quits : Remote host closed the connection |
# 18:27:59 |
bhundven |
ok |
# 18:28:45 |
bhundven |
tlwoerner: so the gcc snapshot for 4.8 (4.8.3) supports arm multilib? |
# 18:32:15 |
bhundven |
tlwoerner: in my point of view, the only horse I have in the multilib race, is just to get it working. |
# 18:33:02 |
bhundven |
tlwoerner: if you have a way to get multilib to work on all architectures that support it, in a way that future architectures will also be supported, I'm all in. |
# 18:34:42 |
bhundven |
tlwoerner: but I'm not interested in a one off change that just gets arm multilib working, disregarding the other platforms. Listing the architectures and variants in ct-ng is planned, we just hadn't gotten there yet. |
# 18:35:01 |
bhundven |
s/Listing/Setting/ |
# 18:51:37 |
mingwandroid |
quits : Read error: Connection reset by peer |
# 18:52:52 |
mingwandroid |
joins #crosstool-ng |
# 18:53:18 |
diorcety |
joins #crosstool-ng |
# 18:59:10 |
mingwandroid_ |
joins #crosstool-ng |
# 19:02:02 |
mingwandroid |
quits : Ping timeout: 264 seconds |
# 19:43:06 |
Virus_7 |
joins #crosstool-ng |
# 19:45:55 |
mingwandroid_ |
quits : Read error: Connection reset by peer |
# 19:47:18 |
mingwandroid |
joins #crosstool-ng |
# 19:52:01 |
mingwandroid_ |
joins #crosstool-ng |
# 19:54:29 |
mingwandroid |
quits : Ping timeout: 240 seconds |
# 19:58:01 |
tlwoerner |
ha! that ctngbot is cute! |
# 19:59:16 |
tlwoerner |
bhundven: if you download the sources from https://launchpad.net/gcc-arm-embedded the compiler declares itself as 4.8.3 and supports arm multilib |
# 19:59:32 |
tlwoerner |
(at least for the cortex-M profile) |
# 19:59:55 |
tlwoerner |
arm-none-eabi-gcc -print-multi-lib |
# 19:59:55 |
tlwoerner |
.; |
# 19:59:55 |
tlwoerner |
thumb;@mthumb |
# 19:59:55 |
tlwoerner |
fpu;@mfloat-abi=hard |
# 19:59:55 |
tlwoerner |
armv6-m;@mthumb@march=armv6s-m |
# 19:59:56 |
tlwoerner |
armv7-m;@mthumb@march=armv7-m |
# 19:59:58 |
tlwoerner |
armv7e-m;@mthumb@march=armv7e-m |
# 20:00:00 |
tlwoerner |
armv7-ar/thumb;@mthumb@march=armv7 |
# 20:00:02 |
tlwoerner |
armv7e-m/softfp;@mthumb@march=armv7e-m@mfloat-abi=softfp@mfpu=fpv4-sp-d16 |
# 20:00:06 |
tlwoerner |
armv7e-m/fpu;@mthumb@march=armv7e-m@mfloat-abi=hard@mfpu=fpv4-sp-d16 |
# 20:00:08 |
tlwoerner |
armv7-ar/thumb/softfp;@mthumb@march=armv7@mfloat-abi=softfp@mfpu=vfpv3-d16 |
# 20:00:10 |
tlwoerner |
armv7-ar/thumb/fpu;@mthumb@march=armv7@mfloat-abi=hard@mfpu=vfpv3-d16 |
# 20:00:56 |
bhundven |
ok |
# 20:01:26 |
bhundven |
so this is work that will go into gcc 4.8.3 (official)? |
# 20:02:21 |
tlwoerner |
i know that team works on getting things upstream, but when it gets in isn't well known |
# 20:02:43 |
tlwoerner |
we would have to try to involve some of them in these discussions |
# 20:02:53 |
tlwoerner |
i've noticed some of them are on the crosstool-NG mailing list |
# 20:03:00 |
bhundven |
yup |
# 20:03:09 |
bhundven |
here is the problem |
# 20:03:17 |
bhundven |
(I've been in your spot before ;) ) |
# 20:03:50 |
bhundven |
crosstool-ng needs to support the official gcc. |
# 20:04:02 |
bhundven |
I also see it working for snapshots |
# 20:04:24 |
bhundven |
but that is more of a gcc/binutils/glibc developer thing |
# 20:05:11 |
bhundven |
I used to work for watchguard, and we did some custom stuff with our toolchain |
# 20:05:42 |
bhundven |
one of them was static toolchain (the tools are static, not that it only makes static output) |
# 20:06:20 |
bhundven |
there has to be a compromise for the gcc-arm-embedded stuff though |
# 20:06:33 |
tlwoerner |
http://en.wikipedia.org/wiki/WatchGuard ? |
# 20:06:53 |
bhundven |
right |
# 20:08:32 |
bhundven |
yann wasn't too excited to take on some of our custom additions to ct-ng |
# 20:08:50 |
bhundven |
so we had to have a local branch to track upstream and our changes |
# 20:09:09 |
tlwoerner |
even if the "C compiler" menu had a generic "extra arguments to pass to pass 1/2/3 build" (i.e. "extra config" the way, say, gdb does) |
# 20:09:22 |
bhundven |
sure |
# 20:09:34 |
bhundven |
I don't see anything wrong with that |
# 20:09:36 |
tlwoerner |
this would allow me to enable "multilib" then pass the "--multilib-list" option |
# 20:09:52 |
bhundven |
but |
# 20:10:46 |
bhundven |
crosstool-ng doesn't really track passes that well |
# 20:11:01 |
bhundven |
well, let me explain |
# 20:12:13 |
tlwoerner |
do_cc_core_pass_1(), do_cc_core_pass_2(), do_cc_for_host() ? |
# 20:12:27 |
bhundven |
hehe |
# 20:12:30 |
bhundven |
yes |
# 20:12:49 |
bhundven |
at the same time as multilib is going in |
# 20:13:03 |
bhundven |
so is llvm/clang |
# 20:13:18 |
bhundven |
and a few other things |
# 20:13:42 |
bhundven |
I just feel like it's about time some refactoring happens to take these things into consideration |
# 20:13:57 |
bhundven |
probably won't be doing multilib for llvm |
# 20:14:08 |
bhundven |
so you're right |
# 20:14:38 |
bhundven |
putting in a CT_CC_GCC_PASS{x}_EXTRA_ARGS isn't going to hurt much |
# 20:14:49 |
tlwoerner |
:-) |
# 20:15:12 |
bhundven |
I've been sick, so my brain is kinda scrambled :P |
# 20:15:21 |
tlwoerner |
i'm not familiar enough with llvm, but it's on "the list" :-) |
# 20:15:42 |
Virus_7 |
quits : Quit: KVIrc 4.2.0 Equilibrium http://www.kvirc.net/ |
# 20:16:09 |
bhundven |
it just seems to me that crosstool-ng needs to be componentized a bit more |
# 20:16:36 |
bhundven |
parts of crosstool-NG.sh.in reach into scripts/build/cc/gcc.sh |
# 20:16:45 |
bhundven |
and have no barring on llvm |
# 20:16:55 |
bhundven |
stuff like that |
# 20:17:02 |
tlwoerner |
well, now that i'm getting more familiar with it, i'd be happy to help, and help with any refactoring (if i understand what i'm doing!) |
# 20:17:16 |
bhundven |
does that make sense? |
# 20:17:44 |
tlwoerner |
oh yes, everything you've said is clear. and i'm grateful for the explanations |
# 20:18:07 |
bhundven |
cool. I feel a bit foggy |
# 20:18:12 |
bhundven |
hehe |
# 20:18:22 |
tlwoerner |
since i have your attention... |
# 20:19:24 |
mingwandroid_ |
quits : Read error: Connection reset by peer |
# 20:20:07 |
tlwoerner |
when one uses a crosscompiler, often there are a bunch of flags that are always given (-ffunction-sections -fdata-sections) |
# 20:20:25 |
tlwoerner |
isn't there a way to build the crosscompiler such that it will always invoke these flags automatically? |
# 20:20:40 |
mingwandroid |
joins #crosstool-ng |
# 20:20:51 |
bhundven |
well, yes. luckly gcc isn't monolithic |
# 20:21:08 |
tlwoerner |
that way the Makefiles and build invocations can be "cleaner"? |
# 20:21:35 |
bhundven |
it depeneds |
# 20:21:48 |
bhundven |
s/depeneds/depends/ |
# 20:22:20 |
bhundven |
so, there are flags you want gcc built with, and there are flags you want gcc to use when compiling on the target |
# 20:22:33 |
bhundven |
er, compiling for the target |
# 20:22:49 |
bhundven |
I'm assuming you mean the later |
# 20:23:21 |
bhundven |
one way is to make a wrapper for the cross-compiler |
# 20:23:46 |
bhundven |
for instance if you have: arm-unknown-linux-gnueabi- |
# 20:24:07 |
tlwoerner |
ah good, _for_ the target |
# 20:24:14 |
tlwoerner |
yes, the latter |
# 20:24:40 |
bhundven |
mv the -gcc to -gcc.orig, and have the wrapper add the command line (this in my opinion is the cleaner way) |
# 20:24:55 |
bhundven |
or |
# 20:25:01 |
tlwoerner |
oh yea? that seems "hackish" to me :-) |
# 20:25:11 |
bhundven |
waits for everyone to get ready to hit him with thousands of shoes |
# 20:25:20 |
bhundven |
you can modify the specs file |
# 20:25:28 |
bhundven |
ducks |
# 20:25:38 |
tlwoerner |
that requires a source patch before building? |
# 20:26:02 |
bhundven |
right |
# 20:26:04 |
tlwoerner |
what about "CFLAGS_FOR_TARGET"? |
# 20:26:35 |
tlwoerner |
everyone's been quiet up to now, if they get excited they're too late to the conversation :-) |
# 20:27:25 |
bhundven |
there was something about that... |
# 20:27:28 |
bhundven |
I forget now |
# 20:27:35 |
bhundven |
let me look really quick |
# 20:28:25 |
tlwoerner |
i agree, modifying a specs file would be much messier than a wrapper |
# 20:28:39 |
bhundven |
well, it's also kinda messy |
# 20:28:48 |
bhundven |
in a different way |
# 20:29:11 |
bhundven |
because rather then modifying the specs file for the build |
# 20:29:20 |
bhundven |
you want to replace the specs file after the build |
# 20:29:32 |
bhundven |
so I was wrong earlier, gcc wouldn't need a patch before building |
# 20:30:02 |
tlwoerner |
oh phew, i was just about to ask (i was getting confused) |
# 20:30:04 |
bhundven |
ct-ng or (manually by hand) you modify the specs file after gcc-final is done |
# 20:30:18 |
bhundven |
yea, sorry about that |
# 20:30:30 |
tlwoerner |
but i thought the specs file was hidden away since a long time ago? |
# 20:30:46 |
tlwoerner |
back in the good ol' days, the specs file was readily available as part of the install |
# 20:30:58 |
tlwoerner |
you'd just copy it, edit it, and replace |
# 20:31:09 |
tlwoerner |
but i though that was gone and the specs were "burried" |
# 20:31:38 |
tlwoerner |
s/were/are |
# 20:33:14 |
bhundven |
nah |
# 20:33:51 |
bhundven |
if a specs file does not exist in the right place |
# 20:34:01 |
bhundven |
it uses the internal specs file |
# 20:34:09 |
bhundven |
-gcc -dumpspecs |
# 20:34:13 |
tlwoerner |
ah, interesting |
# 20:34:31 |
tlwoerner |
how does one know the "right" place? |
# 20:34:34 |
bhundven |
you can dump that to a file and put it in ..\libgcc\specs |
# 20:34:54 |
tlwoerner |
excellent |
# 20:35:08 |
bhundven |
(nasty hack: http://www.linuxfromscratch.org/lfs/view/6.2/chapter05/adjusting.html) |
# 20:35:27 |
bhundven |
we did have some specs file magic at WG |
# 20:35:35 |
bhundven |
and I think that was in with the changes yann didn't want |
# 20:35:46 |
bhundven |
as it is a nasty hack |
# 20:36:02 |
bhundven |
essentially, the specs file is the driver for gcc |
# 20:36:23 |
bhundven |
to tell gcc the steps and what to pass the steps on the command line |
# 20:36:31 |
bhundven |
internally |
# 20:36:47 |
bhundven |
http://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html |
# 20:37:16 |
bhundven |
mingwandroid: doesn't mingw have some specs file magic? |
# 20:37:53 |
mingwandroid |
I'm playing catch up here .. one sec. |
# 20:38:38 |
mingwandroid |
tlwoerner: trevor? hello. |
# 20:38:44 |
tlwoerner |
mingwandroid: hello |
# 20:39:50 |
bhundven |
hehe, my brain had to go back to 2011 for the specs file memory |
# 20:39:56 |
mingwandroid |
quits : Read error: Connection reset by peer |
# 20:40:17 |
bhundven |
think I was on gcc-4.4 |
# 20:41:08 |
mingwandroid |
joins #crosstool-ng |
# 20:41:32 |
mingwandroid |
bhundven: I don't *think* mingw is specifically hacky specs-wise. |
# 20:42:00 |
tlwoerner |
hmm... according to "The Definitive Guide to GCC" the specs file was internalized starting with 4.0 |
# 20:42:13 |
tlwoerner |
it suggests dumping the specs (as you mentioned) |
# 20:42:26 |
tlwoerner |
then using the "-specs=file" cmdline option |
# 20:43:07 |
mingwandroid |
tlwoerner: I missed some context here, what are you trying to achieve? |
# 20:43:24 |
bhundven |
I personally recommend the wrapper option |
# 20:44:04 |
tlwoerner |
internalizing options such as "-ffunction-sections" so the user isn't required to specify them every time they invoke the compiler |
# 20:44:18 |
tlwoerner |
"default arguments" to the crosscompiler... if you will |
# 20:45:00 |
tlwoerner |
which could also be used to specify a linker script too, for different targets |
# 20:45:51 |
bhundven |
ah, here is the more modern stuff from lfs: http://www.linuxfromscratch.org/lfs/view/7.4/chapter06/adjusting.html |
# 20:47:02 |
mingwandroid |
tlwoerner: ah ok. +1 for wrapper scripts from me then. |
# 20:47:13 |
bhundven |
so, I'm happy to hand you the rope to hang yourself, but I'd rather send you to a path of sanity |
# 20:47:20 |
bhundven |
the wrapper is the better choice |
# 20:47:43 |
tlwoerner |
excellent, i have some reading ahead of me then :-D |
# 20:47:55 |
tlwoerner |
yes. i'm starting to agree => wrappers are cleaner |
# 20:47:57 |
bhundven |
hehe |
# 20:47:58 |
mingwandroid |
tlwoerner: linaro has it's own fork of crosstool-ng? |
# 20:48:26 |
tlwoerner |
uhhhh.... not that i know of... (but i'm not in the toolchain group) |
# 20:49:19 |
bhundven |
I know michael hope has done some work on ct-ng |
# 20:49:26 |
bhundven |
(spelling?) |
# 20:49:45 |
bhundven |
but, idk if he has a linaro fork |
# 20:50:10 |
tlwoerner |
yes, i've noticed his contributions before, but i have yet to meet him in person |
# 20:50:31 |
tlwoerner |
linaro has their own gcc fork, which ct-ng can be used to build |
# 20:50:54 |
mingwandroid |
quits : Read error: Connection reset by peer |
# 20:51:34 |
tlwoerner |
i know some of the guys use https://wiki.linaro.org/Cbuildv2Usage |
# 20:52:09 |
mingwandroid |
joins #crosstool-ng |
# 20:52:21 |
bhundven |
So, that is kind of the point I'm trying to get to. This is my thought. (I was waiting for mingwandroid to join again) |
# 20:52:47 |
bhundven |
if scripts/build/cc/gcc.sh is componentized |
# 20:53:05 |
mingwandroid |
diorcety: ping |
# 20:53:25 |
bhundven |
tlwoerner: so that there are wrapper sh files for: official, linaro, custom |
# 20:53:36 |
bhundven |
then there is no need to fork |
# 20:54:04 |
bhundven |
is all about abstracting process, to make more flexable |
# 20:55:11 |
bhundven |
just like breaking out architecture specific things to scripts/build/arch/ |
# 20:55:55 |
mingwandroid |
bhundven: well, I think once we get multilib committed, working on diorcety fork to get the CC-variant stuff into a mergable state would be really good. |
# 20:56:09 |
bhundven |
ja |
# 20:56:41 |
mingwandroid |
where are we with multilib? tlwoerner: that was a long list of multilibs for your compiler. |
# 20:57:51 |
bhundven |
well, I know there are other updates in patchworks that should get applied soon |
# 20:58:01 |
bhundven |
(not my kconfig update, though) |
# 20:58:19 |
bhundven |
but they are mostly version updates |
# 20:59:11 |
bhundven |
I don't think the multilib changes should affect llvm |
# 20:59:21 |
bhundven |
or am I totally wrong? |
# 21:00:10 |
bhundven |
as multilib should be specific to gcc/target-glibc/target-binutils |
# 21:00:53 |
tlwoerner |
mingwandroid: yes, that nice list is generated from the gcc-arm-embedded compiler, which is being developed by ARM employees (https://launchpad.net/gcc-arm-embedded) |
# 21:01:24 |
bhundven |
nods |
# 21:01:39 |
mingwandroid |
bhundven: well, it's not a dependency at all, just thinking "lets get that out out of the way first" |
# 21:02:03 |
tlwoerner |
sounds like there's a number of interesting TODO items for crosstool-NG yet! |
# 21:02:06 |
bhundven |
diorcety's changes are probably 10x more mature then the multilib stuff |
# 21:02:17 |
bhundven |
tlwoerner: oh yea |
# 21:02:31 |
bhundven |
:D |
# 21:02:53 |
mingwandroid |
bhundven: well, I think multilib is quite un-intrusive and can be committed if we can ensuere its not effecting the non-multilib case .. |
# 21:03:02 |
bhundven |
I don't have a personal agenda, other then fielding support on crossgcc and working on my silly projects |
# 21:04:42 |
bhundven |
hopefully, I'm able to help yann a bit more with patchworks and moving to git, so that more people don't have to go through the mercurial learning curve anymore |
# 21:05:19 |
mingwandroid |
bhundven: then CT_EXPERIMENTAL it .. I know there was an attempt by Yann D. to merge diorcety's CC choice stuff .. https://sourceware.org/ml/crossgcc/2013-12/msg00003.html |
# 21:05:59 |
bhundven |
yes, the changes just need to be broken out into steps |
# 21:06:11 |
bhundven |
I haven't looked at it recently |
# 21:06:44 |
mingwandroid |
https://sourceware.org/ml/crossgcc/2013-12/msg00018.html |
# 21:06:50 |
bhundven |
at the same time, we're trying to remove kconfig from ct-ng |
# 21:07:03 |
bhundven |
and switching to y_morin's kconfig-frontends |
# 21:07:06 |
bhundven |
external dependency |
# 21:07:18 |
bhundven |
http://ymorin.is-a-geek.org/git/kconfig-frontends/ |
# 21:07:32 |
bhundven |
(which I have a few pending patches to) |
# 21:07:47 |
mingwandroid |
what's this all about? |
# 21:08:35 |
bhundven |
I posted an update to sync kconfig from the upstream in the linux kernel |
# 21:09:06 |
bhundven |
and yann suggested that we attempt to remove kconfig from ct-ng to make ct-ng a little easier to maintain |
# 21:09:14 |
y_morin |
bhundven: Well, we can keep our built-in version (and bump it once more). But if kconfig stuff already exists on the user's machine, use that and don;t bother build our bundled version. |
# 21:09:23 |
bhundven |
right |
# 21:09:45 |
bhundven |
I'd like to just see it go away, and use the external dep |
# 21:09:47 |
y_morin |
bhundven: Then after a while, we can just rip it out (but not right now). |
# 21:09:52 |
bhundven |
:) |
# 21:09:54 |
y_morin |
bhundven: So would I. |
# 21:10:11 |
bhundven |
jinx |
# 21:14:22 |
mingwandroid |
quits : Read error: Connection reset by peer |
# 21:15:35 |
mingwandroid |
joins #crosstool-ng |
# 21:16:16 |
bhundven |
mingwandroid: I personally think that because no one has really commented on the patches in the patchqueue, it is less mature then llvm/diorcety changes |
# 21:16:40 |
bhundven |
I feel less comfortable pushing changes that aren't complete or commented on |
# 21:17:31 |
mingwandroid |
quits : Read error: Connection reset by peer |
# 21:18:45 |
mingwandroid |
joins #crosstool-ng |
# 21:22:54 |
mingwandroid |
bhundven: I think posting the to the ML is the accepted way to request for comments though is it not? |
# 21:23:13 |
bhundven |
sure |
# 21:23:50 |
bhundven |
tlwoerner: do you want to be added to the ml patch queue to add your additions? |
# 21:24:15 |
tlwoerner |
putting patches on ML will help their visibility |
# 21:24:23 |
tlwoerner |
bhundven: there's a ml patch queue? :-) |
# 21:24:46 |
bhundven |
https://bitbucket.org/bhundven/crosstool-ng-multilib |
# 21:24:56 |
mingwandroid |
ml == mailing list and multilib?! |
# 21:25:05 |
bhundven |
hahaha |
# 21:25:11 |
tlwoerner |
mingwandroid: haha, yes i was thinking that! |
# 21:25:21 |
mingwandroid |
ok ML == mailing list, ml == multilib, not confusing at all now. |
# 21:25:38 |
tlwoerner |
bhundven: i'm not sure if i need to be added |
# 21:25:48 |
mingwandroid |
tlwoerner: have you done some tests of it yet? |
# 21:26:02 |
mingwandroid |
y_morin: would you rather we posted to the mailing list before commenting or would you check that URL ^? |
# 21:26:24 |
bhundven |
there is quite a bit of backlog too |
# 21:26:31 |
bhundven |
on the mailinglist |
# 21:26:42 |
tlwoerner |
bhundven: i did a quick test, but it required me to hard-code my --with-multilib-list, and it failed in the very last step (cleaning up) |
# 21:26:43 |
y_morin |
bhundven: The list, definitely. That way, all can read and review. |
# 21:27:28 |
y_morin |
bhundven: Yes, I've been slacking a bit. But I've already quickly skimmed over all the changes, and most are not urgent. |
# 21:27:33 |
tlwoerner |
but i haven't had time to investigate |
# 21:27:39 |
bhundven |
:) |
# 21:28:01 |
bhundven |
tlwoerner: is that with the patches mingwandroid and I have or with your pass1/2/3 option? |
# 21:28:01 |
mingwandroid |
y_morin: whens 1.20 planned for? |
# 21:28:06 |
y_morin |
bhundven: I want to do the release *this* WE, so I'm building all the toolchains (takes time, so no patch jungling in the meantime. |
# 21:28:15 |
y_morin |
mingwandroid: ^^^ |
# 21:28:19 |
bhundven |
y_morin: understood ;) |
# 21:28:29 |
tlwoerner |
y_morin: the "newlib custom extract" patch is urgent to me :-) |
# 21:28:46 |
y_morin |
tlwoerner: Yes, I saw that one (but it is very recent!) |
# 21:29:11 |
mingwandroid |
so codyps said that newlib {doesn't need to be|isn't} built for multilib? |
# 21:29:11 |
tlwoerner |
it's just a one character change |
# 21:29:28 |
tlwoerner |
mingwandroid: newlib is being built for multilib |
# 21:29:33 |
y_morin |
tlwoerner: We were om Buildroot-release spree the past few days, so I focused on that... |
# 21:29:45 |
mingwandroid |
tlwoerner: with different CFLAGS per each variant? |
# 21:29:45 |
tlwoerner |
mingwandroid: but newlib is smarter and doesn't need all the fixups that {e}glibc needs |
# 21:30:03 |
mingwandroid |
tlwoerner: does it build for some lowest common denominator? |
# 21:31:06 |
tlwoerner |
mingwandroid: how would i tell? by looking at the build log? what i know is that when it is done there are multiple crt0.o files in multiple architecture directories :-) |
# 21:31:35 |
mingwandroid |
not crt0.o, that's core GCC, more libm.a for example. |
# 21:32:03 |
mingwandroid |
GCC will definitely do a per-multilib build, but in a different folder layout than the sysroot layout, annoyingly. |
# 21:33:34 |
tlwoerner |
./arm-unknown-eabi/sysroot/lib/armv6-m/libm.a |
# 21:33:34 |
tlwoerner |
./arm-unknown-eabi/sysroot/lib/libm.a |
# 21:33:34 |
tlwoerner |
./arm-unknown-eabi/sysroot/lib/thumb/libm.a |
# 21:33:34 |
tlwoerner |
./arm-unknown-eabi/sysroot/lib/armv7e-m/libm.a |
# 21:33:34 |
tlwoerner |
./arm-unknown-eabi/sysroot/lib/armv7e-m/fpu/libm.a |
# 21:33:35 |
tlwoerner |
./arm-unknown-eabi/sysroot/lib/armv7e-m/softfp/libm.a |
# 21:33:37 |
tlwoerner |
./arm-unknown-eabi/sysroot/lib/armv7-m/libm.a |
# 21:33:39 |
tlwoerner |
./arm-unknown-eabi/sysroot/lib/fpu/libm.a |
# 21:33:43 |
tlwoerner |
./arm-unknown-eabi/sysroot/lib/armv7-ar/thumb/libm.a |
# 21:33:45 |
tlwoerner |
./arm-unknown-eabi/sysroot/lib/armv7-ar/thumb/fpu/libm.a |
# 21:33:47 |
tlwoerner |
./arm-unknown-eabi/sysroot/lib/armv7-ar/thumb/softfp/libm.a |
# 21:34:13 |
bhundven |
hears y_morin pass out |
# 21:34:17 |
y_morin |
tlwoerner: Please use a pastebin site to paste more than 3-4 lines... |
# 21:34:21 |
y_morin |
bhundven: ;-) |
# 21:34:34 |
tlwoerner |
haha, sorry |
# 21:34:41 |
y_morin |
bhundven: You know me well! ;-) |
# 21:34:48 |
y_morin |
tlwoerner: no problem. ;-) |
# 21:38:44 |
tlwoerner |
mingwandroid: yes, there are per-architecture flags for each multilib when building newlib |
# 21:38:50 |
tlwoerner |
(looking through the log file) |
# 21:38:59 |
tlwoerner |
i'll post it... would here be okay? ;-) |
# 21:39:29 |
mingwandroid |
tlwoerner: ok. It's many years since I dealt with newlib (apart from in Cygwin/MSYS2) .. |
# 21:39:51 |
mingwandroid |
does newlib ask GCC for the details then? |
# 21:41:26 |
tlwoerner |
mingwandroid: i'm new to being at this level myself :-) |
# 21:41:38 |
tlwoerner |
i'm just preparing the newlib part of the build log so you can have a look |
# 21:44:28 |
bhundven |
tlwoerner: personally, I'm really excited to see newlib+multilib+cortex-m{0,1,2,3...} |
# 21:44:42 |
bhundven |
I have a few of those |
# 21:45:14 |
bhundven |
I was just trying to handle some of the easier (e)glibc variants/architectures before getting to arm |
# 21:45:29 |
bhundven |
and newlib was an unknown to me |
# 21:45:41 |
bhundven |
and we've also been talking about mucl |
# 21:45:42 |
tlwoerner |
instead of pastebin, how does one share a file? |
# 21:45:55 |
mingwandroid |
paste.kde.org for me. |
# 21:46:07 |
bhundven |
less limit on size |
# 21:46:58 |
tlwoerner |
but again, that requires me to cut and past the whole file, how do i just upload the file? |
# 21:47:15 |
mingwandroid |
tlwoerner: dropbox is an option? |
# 21:47:22 |
bhundven |
google drive |
# 21:47:28 |
bhundven |
mediafire |
# 21:47:41 |
mingwandroid |
does google drive support linux yet? |
# 21:47:48 |
bhundven |
ugh, kinda |
# 21:47:54 |
bhundven |
well, not by google |
# 21:47:55 |
bhundven |
:) |
# 21:48:38 |
mingwandroid |
ffs. come on google! |
# 21:48:41 |
tlwoerner |
linux is all i use, and we use google drive all the time (the linaro email system is actually a corporate gmail account) |
# 21:49:26 |
mingwandroid |
tlwoerner: well, I'm asking about a client like dropbox .. shell integration for popular file managers, the usual nicities that DB give us |
# 21:49:47 |
bhundven |
there are some different fusefs addons |
# 21:49:53 |
tlwoerner |
ah, no. just web interface |
# 21:54:03 |
tlwoerner |
https://drive.google.com/file/d/0B3mtE9WJovhfSGI0M0ZreUdyV2M/edit?usp=sharing |
# 21:54:20 |
tlwoerner |
man... you guys are fast! |
# 21:55:04 |
tlwoerner |
what you must remember is that the sources for this build come from the gcc-arm-embedded project |
# 21:55:16 |
tlwoerner |
so i'm not sure if they've done anything to help things along |
# 21:55:21 |
bhundven |
I may be fast, but my connection is teh slowz |
# 21:55:23 |
tlwoerner |
obviously they've done some multilib work |
# 21:59:38 |
bhundven |
wholy smokes |
# 21:59:51 |
tlwoerner |
bhundven: i agree, i'm looking forward to doing much more cortex-M work too, and i've got a couple boards lined up |
# 22:00:29 |
mingwandroid |
quits : Read error: Connection reset by peer |
# 22:01:08 |
bhundven |
that is a serious ./configure line |
# 22:01:49 |
mingwandroid |
joins #crosstool-ng |
# 22:02:01 |
mingwandroid |
[ALL ] Checking multilib configuration for newlib... |
# 22:02:01 |
mingwandroid |
[ALL ] Checking multilib configuration for libgloss... |
# 22:02:11 |
bhundven |
looks at his coffee, thinks he should be on crack |
# 22:02:56 |
mingwandroid |
[ALL ] Adding multilib support to Makefile in /home/trevor/devel/arm/toolchains/crosstool-ng/mercurial/work-area/newlib-test/_build/src/newlib-custom/libgloss |
# 22:03:18 |
mingwandroid |
.. so yeah, newlib does the stuff we've added by the look of it .. clever newlib, clap clap. |
# 22:03:29 |
bhundven |
hmm |
# 22:04:20 |
tlwoerner |
from the sounds of it, the gcc-arm-embedded people have pretty much taken over maintainership of newlib |
# 22:04:33 |
tlwoerner |
they even have their own "newlib-nano" they're working on too |
# 22:04:37 |
tlwoerner |
https://answers.launchpad.net/gcc-arm-embedded/+question/244125 |
# 22:06:13 |
bhundven |
bleh, not open though |
# 22:06:49 |
tlwoerner |
but the sources are available (for each release) |
# 22:07:03 |
bhundven |
I wish linaro would work on (e)glibc multilib... |
# 22:07:08 |
bhundven |
hehe |
# 22:07:30 |
bhundven |
make it that easy |
# 22:07:34 |
bhundven |
er as newlib |
# 22:07:59 |
tlwoerner |
bhundven: if you find a customer willing to pay for it, linaro will do anything :-) |
# 22:08:17 |
bhundven |
looks in his pocket, pulls out $2 USD |
# 22:08:22 |
bhundven |
:( |
# 22:08:33 |
bhundven |
lol |
# 22:08:59 |
tlwoerner |
:-D |
# 22:09:48 |
bhundven |
maybe when this startup pays off... |
# 22:10:01 |
bhundven |
s/when/if/ |
# 22:16:59 |
tlwoerner |
i believe the gcc-arm-toolchain people push their code upstream, so this work is being done (at least for the cortex-M case) |
# 22:17:10 |
tlwoerner |
linaro tends to focus on the cortex-A space |
# 22:28:25 |
smartin_ |
joins #crosstool-ng |
# 22:51:34 |
bhundven |
tlwoerner: I'm really happy with this efm32 dk3750gg http://www.silabs.com/products/mcu/lowpower/Pages/efm32gg-dk3750.aspx |
# 22:51:47 |
bhundven |
fun project board |
# 22:53:36 |
bhundven |
these guys are trying to port linux to it: http://lwn.net/Articles/476996/ |
# 23:01:43 |
sh4rm4 |
gotta love arm boards where you need special kernel support for each of them |
# 23:01:53 |
sh4rm4 |
bios ftw |
# 23:02:03 |
bhundven |
hehe |
# 23:05:32 |
bhundven |
sh4rm4: and in turn, gotta love arm fragmentation |
# 23:05:57 |
bhundven |
not saying energymicro/siliconlabs is helping there. |
# 23:06:37 |
sh4rm4 |
honestly i wouldnt have bought my arm netbook had i known earlier how much hacking it needs |
# 23:07:01 |
sh4rm4 |
and that i need to use an old unmaintained 2.6.31 kernel with a ton of custom patches |
# 23:07:12 |
bhundven |
which netbook? |
# 23:07:15 |
sh4rm4 |
efika mx |
# 23:07:51 |
bhundven |
seems like a good candidate for chromeos |
# 23:08:10 |
sh4rm4 |
just that half of the drivers didnt go upstream |
# 23:08:15 |
bhundven |
ah |
# 23:08:22 |
sh4rm4 |
so youre stuck with the olde kernel |
# 23:08:45 |
bhundven |
well, it's a good spot to be in, if you want to become a kernel contributor |
# 23:08:46 |
sh4rm4 |
unless you have a lot of time to port patches to recent |
# 23:08:54 |
bhundven |
yup |
# 23:10:10 |
bhundven |
I've given a few stabs at trying to port this s5pc110 android device (SGH-T959V, based on nexus s (crespo)) to newer kernels. it's painful. |
# 23:10:35 |
bhundven |
without jtag anyway |
# 23:11:18 |
bhundven |
we have a 2.6.35.x and 3.0.x kernel, but samsung really did a lame coding job. |
# 23:11:59 |
sh4rm4 |
the only thing my netbook is good at is battery life and noise. but it's dead slow, nearly no ram, buggy kernel, minimal resolution, no ethernet on board, only 16gflash... |
# 23:12:15 |
sh4rm4 |
an atom netbook would've been 10x more useful for the same price |
# 23:12:37 |
sh4rm4 |
compiling webkitgtk took 2.5 days |
# 23:12:38 |
bhundven |
that target (atom) has changed a lot though. started with lpia |
# 23:13:03 |
bhundven |
remembers his Samsung Q1U |
# 23:13:16 |
bhundven |
lost to the pawn shop |
# 23:13:19 |
sh4rm4 |
afaik you can use a vanilla kernel with any atom device |
# 23:13:24 |
bhundven |
yup |
# 23:14:18 |
bhundven |
but for better efficiency, there are some toolchain optimizations you can use. |
# 23:14:45 |
bhundven |
linux pm doesn't take advantage of it though |
# 23:14:56 |
bhundven |
same with the efm32 |
# 23:15:06 |
bhundven |
super low power states |
# 23:19:22 |
bhundven |
I've wanted to get an arm netbook, but I'm waiting for ssd to go down in price and up in capacity |
# 23:19:47 |
sh4rm4 |
well i'd actually prefer a proper drive |
# 23:19:50 |
sh4rm4 |
1 TB or so |
# 23:19:59 |
sh4rm4 |
so you can store a lot of games on it |
# 23:20:17 |
sh4rm4 |
mame roms, nes roms, etc |
# 23:20:52 |
bhundven |
hehe |
# 23:21:05 |
bhundven |
I play nes roms on my psp :D |
# 23:21:13 |
sh4rm4 |
:) |
# 23:47:33 |
smartin_ |
quits : Write error: Broken pipe |