# 02:46:44 |
falstaff |
joins #crosstool-ng |
# 02:47:43 |
falstaff_ |
quits : Ping timeout: 246 seconds |
# 03:03:29 |
bhundven |
joins #crosstool-ng |
# 04:07:56 |
tkil |
joins #crosstool-ng |
# 04:08:34 |
tkil |
anyone cross-compiling 'perf' along with their linux kernel? |
# 04:10:23 |
bhundven |
tkil: there are some known issues on certain architectures |
# 04:10:31 |
bhundven |
which are you using? |
# 04:13:11 |
tkil |
bhundven -- x86-64 host, ppc32 host. |
# 04:13:33 |
tkil |
and my issues don't seem to be host/target related, so much as pure "those headers don't do what you think they do". |
# 04:13:59 |
bhundven |
which kernel version? |
# 04:14:38 |
tkil |
http://lkml.indiana.edu/hypermail/linux/kernel/1306.3/04254.html and http://paste.stg.fedoraproject.org/4255/54140813/ |
# 04:14:40 |
tkil |
3.9.7 |
# 04:14:54 |
tkil |
SFAICT, it happened with the "UAPI" header refactoring, which hit in 3.7 or so. |
# 04:15:11 |
tkil |
er, x86-64 host, ppc32 target. sigh. |
# 04:17:31 |
bhundven |
yikes |
# 04:17:36 |
tkil |
i have scripts that worked fine with 2.6.late, 3.0.x, and 3.4.x |
# 04:17:42 |
tkil |
yeah. :) |
# 04:17:54 |
tkil |
and i can't get away with plain "-nostdinc", since perf uses stdio.h and friends... |
# 04:18:00 |
bhundven |
well, considering that anthony posted that just a little bit ago |
# 04:18:09 |
tkil |
oops. that's me. |
# 04:18:14 |
bhundven |
;) |
# 04:18:21 |
bhundven |
didn't recognize the nick |
# 04:18:27 |
tkil |
it's in the email addr, i think. |
# 04:18:37 |
tkil |
unintentional alt, i've had this nick since 1986 or so. |
# 04:18:51 |
bhundven |
hehe, shows how much I pay attention |
# 04:18:52 |
tkil |
==anthony.foiani@gmail.com, which i use on the list. sorry. |
# 04:19:08 |
bhundven |
well, also considering |
# 04:19:20 |
tkil |
heh. how dare you not memorize all the names. (i think we had two Yanns on the list in the last 3-4mo, IIRC...) |
# 04:19:41 |
bhundven |
that there has been a lot of focus on perf by linaro and the kernel community, I wouldn't be surprised to see a response soon. |
# 04:20:30 |
bhundven |
yea, I'm glad the bryan hundven's of the world are pretty rare. |
# 04:21:33 |
bhundven |
although, you might have better anonymity as a John Smith :) |
# 04:21:55 |
tkil |
heh. |
# 04:22:04 |
tkil |
not too many "foianis" in the world, either. |
# 04:22:14 |
bhundven |
well, "pre prism" |
# 04:22:25 |
bhundven |
lol |
# 04:22:40 |
bhundven |
yea |
# 04:22:49 |
tkil |
cool, i hope i hear back soon. |
# 04:22:56 |
tkil |
driving me a little insane. |
# 04:23:45 |
bhundven |
I miss working with embedded linux as much as I used to. I've been mostly doing java lately. bleh. |
# 04:25:21 |
tkil |
i worked the java mines in the mid 00s... dunno if i would prefer the "it's broken, but at least it's *defined* to be broken" vs. "something broke, is it yours, the compiler's, the kernel's, ctng's..." |
# 04:26:31 |
bhundven |
I like the $2 of FS="vs.". More challange. Different languages that were designed for specific things. |
# 04:27:26 |
bhundven |
kind of like playing with logic puzzles |
# 04:27:33 |
tkil |
java gives you some of that, too. the magical log4j init files, and J2EE container properties... |
# 04:28:00 |
bhundven |
I guess. |
# 04:28:16 |
bhundven |
maybe it just doesn't seem that fun to me. |
# 04:28:25 |
tkil |
:) just trying to think of the corresponding type of thing that annoyed me when i was doing java... |
# 04:29:09 |
tkil |
i mean, getting perf cross-compiled is about 8 steps away from straight-line progress on what i'm actually trying to accomplish... |
# 04:29:10 |
bhundven |
I could write a 40 line python script that does the same thing my 2500 line java program does. |
# 04:29:18 |
tkil |
yeah, it's noisy. |
# 04:29:37 |
tkil |
http://en.wiktionary.org/wiki/yak_shaving |
# 04:29:42 |
bhundven |
lol |
# 04:30:55 |
tkil |
write it in 40 lines of jython? :) |
# 04:31:15 |
bhundven |
android library |
# 04:31:29 |
bhundven |
idk if android has jython |
# 04:32:12 |
bhundven |
although, that would be pretty bad-a** |
# 04:32:48 |
bhundven |
Google: read this ^^^ |
# 04:33:05 |
bhundven |
:) |
# 04:34:47 |
bhundven |
when you work for a startup, you have to wear multiple hats. No one around to write android/ios/winmo apps for this project I wrote firmware for. |
# 04:35:09 |
bhundven |
(for a device that mobile devices connect to) |
# 04:35:29 |
bhundven |
so, next you'll hear me complain about c# |
# 04:35:30 |
bhundven |
:p |
# 04:35:42 |
bhundven |
no complaints about objective-c or xcode |
# 04:35:50 |
bhundven |
I loved that experience and got that part done in 2 days. |
# 04:39:33 |
tkil |
neat. |
# 04:39:50 |
tkil |
my one experience at hardcore startup (4-5 ppl), yeah, lots of hats. |
# 04:40:16 |
tkil |
closest i ever came to that was really post-startup, ~100ppl, still wore multiple hats but not nearly that severely. |
# 04:43:45 |
tkil |
w.r.t. jython, i thought the point was that you'd eventually end up with a .class file, and not have to worry about the impl language? |
# 04:43:58 |
tkil |
or is it more about running some sort of python w/ easy bindings to JRE? |
# 04:44:12 |
tkil |
was a perl nerd, never got around to learning python in depth. |
# 04:45:59 |
bhundven |
oh... |
# 04:46:50 |
bhundven |
https://code.google.com/p/android-scripting/ |
# 04:47:04 |
bhundven |
thats kinda cool |
# 04:48:40 |
bhundven |
but yea, doesn't seem to be any support for jython |
# 04:50:01 |
bhundven |
but I'm not sure, not researched enough on my part |
# 04:50:45 |
tkil |
http://wiki.python.org/jython/JythonFaq/GeneralInfo "It consists of a compiler to compile Python source code down to Java bytecodes which can run directly on a JVM ..." |
# 04:51:10 |
bhundven |
I think I'll K.I.S.S, and stick with normal java&android-api |
# 04:51:10 |
tkil |
ha. searching for "jython android" gets you back to the android-scripting one... |
# 04:51:29 |
bhundven |
yea |
# 04:51:52 |
tkil |
anyway. back to mumbling different incantations of "-nostdinc ...". |
# 04:51:57 |
bhundven |
lol |
# 04:52:00 |
bhundven |
:) |
# 04:53:46 |
tkil |
check this one out: http://paste.stg.fedoraproject.org/4257/25680191/ |
# 04:55:02 |
tkil |
it defines a type, then manages to include something that redefines it. joy. |
# 04:55:13 |
bhundven |
oh man, that was a big change |
# 04:55:56 |
bhundven |
and a lot of reputable acks. |
# 04:56:23 |
tkil |
yeah. my gut feeling is that people just don't cross-compile perf that much. |
# 04:56:29 |
tkil |
as you said, maybe the linaro folks have some input. |
# 04:56:38 |
bhundven |
have you tried bisect-ing to find exactly which change caused it? |
# 04:56:40 |
tkil |
i should probably just disable it and move on for now. |
# 04:56:58 |
tkil |
huh. i suppose i could, but my build system isn't really set up for that. :( |
# 04:57:57 |
bhundven |
you could just write a bisect script that just tries building perf |
# 04:58:41 |
tkil |
except, it's not the perf build itself that is broken; it's the linux headers_install that is broken (SFAICT) |
# 04:59:18 |
tkil |
so i'd have to bisect and do a multi-point rebuild after each bisect (possibly even re-building ct-ng, since it depends on kernel headers) |
# 04:59:45 |
tkil |
even on this [quite beefy] system, that's about an hour per iter even w/o counting writing the scripting itself... |
# 05:00:15 |
bhundven |
hmm. just disable uapi, build your toolchain. Then test the headers-install with bisect |
# 05:00:18 |
tkil |
er, "headers_install that is broken" s/broken/incompatible/, to be more precise. |
# 05:01:25 |
bhundven |
you can use a default config and add CONFIG_ to the make command line |
# 05:01:49 |
bhundven |
to make it easier to script, and no worries about oldconfig |
# 05:02:03 |
tkil |
huh. didn't realize UAPI was toggleable, that would save me for now... |
# 05:02:07 |
tkil |
goes off to read up. |
# 05:02:26 |
bhundven |
is uapi a required? |
# 05:02:52 |
tkil |
SFAIK, it was a commit that changed stuff [using a script, but still, it altered the repo]. |
# 05:02:56 |
tkil |
but i'll go check. |
# 05:05:16 |
tkil |
yes, it was a hard commit to the repo: https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=607ca46e97a1b6594b29647d98a32d545c24bdff |
# 05:05:30 |
bhundven |
yea |
# 05:05:31 |
tkil |
that delta was *generated* by a script [IIRC], but the actual change was a pure textual change. |
# 05:05:33 |
bhundven |
lame |
# 05:05:42 |
bhundven |
oh, ic |
# 05:05:44 |
tkil |
well, it's the right answer, and i don't begrudge them. |
# 05:05:51 |
bhundven |
no doubt |
# 05:06:10 |
tkil |
but this whole "perf is in the kernel tree, so it's synced with kernel data structures, but..." is biting me. :( |
# 05:06:21 |
tkil |
[which you know and understand. sorry, i'm venting.] |
# 05:06:37 |
bhundven |
well, build a toolchain with an older kernel |
# 05:06:48 |
tkil |
oh, i can switch back to 3.4 with a one-line change. |
# 05:06:54 |
bhundven |
then bisect the newer one with the cross compiler, |
# 05:06:56 |
tkil |
but i need/want fixes in the later kernels. :( |
# 05:07:20 |
tkil |
ah, for finding the change. |
# 05:07:33 |
tkil |
bet you dollars to donuts that's the root, though. :) |
# 05:07:35 |
bhundven |
you wouldn't need to build ctng toolchain to do headers-install |
# 05:07:48 |
bhundven |
maybe |
# 05:07:56 |
bhundven |
stranger things have happened |
# 05:08:17 |
tkil |
:) i have a good nose for these things, but maybe i shouldn't jinx it: https://bugzilla.redhat.com/show_bug.cgi?id=928246 |
# 05:09:08 |
bhundven |
heh :) |
# 06:42:12 |
smartin_ |
joins #crosstool-ng |
# 06:58:53 |
Thomas2 |
joins #crosstool-ng |
# 07:01:43 |
Thomas1 |
quits : Ping timeout: 264 seconds |
# 07:23:47 |
doc2 |
joins #crosstool-ng |
# 12:01:23 |
doc2 |
quits : Ping timeout: 256 seconds |
# 12:25:29 |
kos_tom |
hello |
# 12:25:47 |
kos_tom |
has anyone ever seen the following messages when building eglibc: |
# 12:25:48 |
kos_tom |
/home/test/outputs/eglibc/build/eglibc-23400/build/elf/librtld.os: In function `open_verify': |
# 12:25:51 |
kos_tom |
raise.c:(.text+0x4d6c): undefined reference to `rtld_errno' |
# 14:58:31 |
diorcety |
quits : Read error: Connection reset by peer |
# 15:01:06 |
diorcety |
joins #crosstool-ng |
# 17:04:19 |
doc2 |
joins #crosstool-ng |
# 17:13:10 |
y_morin |
joins #crosstool-ng |
# 19:47:49 |
Thomas2 |
quits : Quit: Thomas2 |
# 20:32:14 |
doc2 |
quits : Remote host closed the connection |
# 22:04:08 |
smartin_ |
quits : Quit: good night all |
# 23:05:06 |
y_morin |
quits : Quit: Nighty Night! |