# 03:49:53 |
trigg |
joins #crosstool-ng |
# 03:50:03 |
trigg_ |
quits : Ping timeout: 240 seconds |
# 03:53:32 |
trigg_ |
joins #crosstool-ng |
# 03:54:14 |
trigg |
quits : Ping timeout: 244 seconds |
# 05:54:50 |
diorcety |
joins #crosstool-ng |
# 06:12:24 |
diorcety |
quits : Read error: Connection reset by peer |
# 06:55:42 |
lundmar |
joins #crosstool-ng |
# 08:41:03 |
jyelloz |
quits : Ping timeout: 276 seconds |
# 09:21:25 |
jyelloz |
joins #crosstool-ng |
# 09:49:23 |
ragedragon |
joins #crosstool-ng |
# 10:56:47 |
ragedragon |
quits : Ping timeout: 260 seconds |
# 12:20:54 |
Net147 |
quits : Read error: Connection reset by peer |
# 12:24:48 |
Net147 |
joins #crosstool-ng |
# 12:58:44 |
lundmar1 |
joins #crosstool-ng |
# 12:58:44 |
lundmar |
quits : Read error: Connection reset by peer |
# 14:00:22 |
blueness |
quits : Quit: blueness |
# 14:02:16 |
blueness |
joins #crosstool-ng |
# 14:15:41 |
blueness |
quits : Quit: blueness |
# 14:16:16 |
blueness |
joins #crosstool-ng |
# 14:31:32 |
blueness |
quits : Quit: blueness |
# 14:32:34 |
blueness |
joins #crosstool-ng |
# 14:36:00 |
lundmar1 |
is now known as: lundmar |
# 15:59:11 |
y_morin |
joins #crosstool-ng |
# 16:06:14 |
bhundven |
joins #crosstool-ng |
# 16:06:19 |
bhundven |
y_morin: hello |
# 16:18:10 |
blueness |
quits : Quit: blueness |
# 16:20:41 |
blueness |
joins #crosstool-ng |
# 16:48:53 |
blueness |
quits : Quit: blueness |
# 16:51:26 |
blueness |
joins #crosstool-ng |
# 18:16:12 |
enunes |
quits : Ping timeout: 240 seconds |
# 19:00:53 |
diorcety |
joins #crosstool-ng |
# 20:21:19 |
blueness |
quits : Quit: blueness |
# 20:32:49 |
lundmar |
damnit, ct-ng travis-ci builds faster than my own machine. This is not good ^^ |
# 20:41:28 |
bhundven |
heh |
# 20:42:26 |
bhundven |
y_morin: are you avail? |
# 20:45:44 |
lundmar |
bhundven: just pushed a oneline pull request in your general direction that you probably wanna merge and close ;) |
# 20:48:01 |
y_morin |
bhundven: Hey! I am now. |
# 20:48:22 |
bhundven |
y_morin: hello! :) |
# 20:49:11 |
bhundven |
y_morin: I was wondering (as aneyman__ has...) if/when you would have some cycles to finish the multilib[#383] PR review? |
# 20:49:48 |
y_morin |
bhundven, aneyman__: Not sure when I can look at it again... :-/ |
# 20:50:25 |
bhundven |
y_morin: ok, could you point me to the commit you left off on, so that I can focus on the remaining commits? |
# 20:50:37 |
y_morin |
I'm sorry about that. I find the topic quite complex, and since I'm not directly interested in multiarch/multilib, forcing myself to look at it is not really easy... :-/ |
# 20:50:50 |
bhundven |
y_morin: totally understand |
# 20:51:16 |
y_morin |
bhundven: And TBH, I find using github to do the review to be less than optimum (commits are not even ordered in commit-order...) |
# 20:51:20 |
bhundven |
y_morin: I was hoping that mingwandroid would of had time to review this as well, since he and I wrote the original work |
# 20:51:41 |
bhundven |
y_morin: I also dislike github's review process/tooling |
# 20:52:02 |
lundmar |
reviewable.io ^^ |
# 20:52:08 |
bhundven |
lundmar: I hated it |
# 20:52:10 |
y_morin |
bhundven: I can try to have a look in a few minutes, but I'll simply give an overall "feeling" on the series. |
# 20:52:25 |
bhundven |
y_morin: ok :) |
# 20:53:01 |
bhundven |
lundmar: more specifically, the UI/UX wasn't compatible with my brain :) |
# 20:53:08 |
lundmar |
bhundven: that pull request is the first time I saw reviewable.io and I agree - there is something off about it. |
# 20:53:19 |
blueness |
joins #crosstool-ng |
# 20:53:24 |
lundmar |
reviewninja might be better choice |
# 20:53:33 |
bhundven |
lundmar: thanks. I look into it |
# 20:59:58 |
lundmar |
I also agree with y_morin, multilib is troublesome. It might not be worth the added complexity or maintenance cost. I think most people use ct-ng to specifically optmize their toolchain for each arch so multiarch often becomes a mute point. However, some users obviously finds it useful. |
# 21:00:49 |
ragedragon |
joins #crosstool-ng |
# 21:02:17 |
y_morin |
lundmar: Yes, when I was maintraining it, my goal was to be able to build a toolchain optimised for a specific target. Hence multilib (and later multiarch) were not really into the equation. |
# 21:02:37 |
lundmar |
nods |
# 21:02:49 |
y_morin |
And given the size of storage these days, having a toolchain per target is really a no-brainer. |
# 21:03:04 |
lundmar |
little cost that is yes |
# 21:06:04 |
lundmar |
y_morin: are you still interested in build systems? I remember you once did a note comparing various build tools ;) |
# 21:07:16 |
y_morin |
lundmar: Well, I would not contribute to Buildroot if I were not! ;-) |
# 21:07:21 |
lundmar |
hehe |
# 21:07:40 |
y_morin |
(and that note was really a loooong time ago, before Buildroot is in the shape it is today) |
# 21:07:57 |
lundmar |
I know ^^ |
# 21:08:06 |
lundmar |
y_morin: anyway, you might wanna add http://buildgear.io to the comparison ;) |
# 21:08:30 |
lundmar |
I remember reading your note back then and I couldn't agree more with it hehe |
# 21:08:34 |
aneyman__ |
bhundven, y_morin: well, then stop pretending like ct-ng supports multilib and just drop that config option completely |
# 21:08:48 |
y_morin |
But I found that Buildroot was really looking to go the way I wanted to go, and is going the way I want to go, so I simply ditched the idea I had to come up with yet-another... ;-) |
# 21:08:55 |
aneyman__ |
because currently it breaks more often than works |
# 21:09:03 |
lundmar |
so I kind of made my own take on a simple build tool, buildgear - for fun. |
# 21:09:57 |
lundmar |
y_morin: well, given the choice between buildroot and anything OE based I would pick buildroot any day |
# 21:10:11 |
aneyman__ |
in the current state, it is effectively only supported for baremetal and the flavors of libc that are inherently multilib (think avr-libc) |
# 21:10:35 |
lundmar |
however, both buildroot and OE based is troublesome if you want to do small and quick from scratch with clean meta |
# 21:23:14 |
y_morin |
aneyman__: Hello! I certainly understand your frustration. I'm sorry about that. :-( |
# 21:23:52 |
y_morin |
aneyman__: Fact is, reviewing takes a lot of time, especially on a topic I am not at ease with, and not entirely interested in. |
# 21:25:13 |
y_morin |
aneyman__: I certainly understand that this is something you *are* interested in (or you would probably not have worked so hard on it!). |
# 21:26:02 |
y_morin |
aneyman__: Yet, I'm doing that at home, in my sapre time, with various priorities all ontending for that spare (very spare!) time. |
# 21:27:25 |
y_morin |
Ditto for bhundven; that's also a hobby activity for him. |
# 21:28:18 |
y_morin |
Yes, it takes time for complex stuff to get in (eh, I even have series against Buildroot that I've started three years ago, and are not entirely upstream yet!) |
# 21:28:29 |
aneyman__ |
y_morin: I understand that - but let's be honest, at the current rate it is not going to be ever reviewed |
# 21:29:34 |
y_morin |
aneyman__: Then maybe because not many people are interested in it... (but yes, the ct-ng comunity is rather small...) |
# 21:30:08 |
y_morin |
aneyman__: I'm not sure what to say. I'm really sorry that this has not lived to your expectation, and I can understand the frustration. |
# 21:30:23 |
aneyman__ |
y_morin: so the choice here is (1) to just stop pretending ct-ng has any notion of multilib support (I can understand that - we'll find another way to do that), (2) do a cursory glance with a testing that it does not break anything or (3) speed up the review |
# 21:30:34 |
aneyman__ |
just pick any of the three |
# 21:31:11 |
y_morin |
aneyman__: You are probably right. I just have something else to suggest: |
# 21:32:10 |
y_morin |
aneyman__: When doing complex series, from experience, it was much faster to go with short series of simple changes (e.g your initial changes) rather than a big and long series with a lot of complex patches. |
# 21:32:37 |
y_morin |
aneyman__: A long series get the maintainer and reviewers all go like "Damn, this will be difficult!" |
# 21:32:37 |
aneyman__ |
y_morin: we already went through this discussion |
# 21:32:56 |
aneyman__ |
when I started this multilib work, more than 50% of ct-ng samples were not building |
# 21:32:57 |
y_morin |
aneyman__: While a short series has more chances to get reviewd faster. |
# 21:33:19 |
aneyman__ |
look up the "fix ct-ng samples" branch to get what I am doing |
# 21:33:49 |
aneyman__ |
and even that branch - which didn't bring *ANY* new features, just restoring the status quo - took more than two months to merge |
# 21:34:16 |
aneyman__ |
so what was I supposed to do during that time? of course, that meant that the multilib stuff kept accumulating on another branch |
# 21:35:41 |
y_morin |
I understand... I don't have much solution, unfortunately... :-/ |
# 21:35:43 |
aneyman__ |
and as a cherry on the cake - I already did that pass to restore the samples before - around last fall |
# 21:36:25 |
aneyman__ |
but the merges of other branches just broke them - looks like the merges were not tested beyond from the samples built by travis-ci |
# 21:36:53 |
aneyman__ |
e.g. new gdb, new uclibc, new gcc all broke one or more samples |
# 21:37:12 |
aneyman__ |
I understand that 'ct-ng build-all' takes about 30 hours |
# 21:37:23 |
aneyman__ |
but that's something that can be done at least weekly |
# 21:40:39 |
y_morin |
An I guess travis is there for that, no? |
# 21:41:30 |
aneyman__ |
travis builds only 5 or 6 samples |
# 21:42:26 |
aneyman__ |
ok, it's 9 now |
# 21:42:40 |
y_morin |
Ah, otherwise the build time is too long and the build is killed, right? |
# 21:42:40 |
aneyman__ |
of which 5 are arm-based |
# 21:43:15 |
aneyman__ |
no, it does not get killed as each sample is considered a separate build - so in finishes within the limit |
# 21:44:53 |
y_morin |
bhundven: Any reason why you did not set up a travis build for each samples, and only a subset? |
# 21:45:13 |
bhundven |
build time |
# 21:47:02 |
y_morin |
bhundven: But if each sample is build as its own job, there is no time limit that would fail the build, is there? |
# 21:47:16 |
aneyman__ |
https://travis-ci.org/crosstool-ng/crosstool-ng/builds/86550894 |
# 21:47:23 |
y_morin |
bhundven: E.g. with Buildroot, our travis bot builds all our defconfigs. |
# 21:47:26 |
bhundven |
there is a limit to the number of workers we get |
# 21:47:27 |
aneyman__ |
that was our try to build all :) |
# 21:47:37 |
aneyman__ |
since then, the number of samples doubled |
# 21:47:43 |
bhundven |
more workers means we need to get a paid account |
# 21:47:52 |
bhundven |
I don't have the resources for that |
# 21:48:07 |
aneyman__ |
well maybe not doubled but still increased |
# 21:48:34 |
bhundven |
I try to get a cross-section of samples that exercise ct-ng's features |
# 21:49:00 |
bhundven |
while also not making contributors wait 2 days to test a PR |
# 21:49:30 |
bhundven |
(or to wait a day to find out they misspelled a variable name) |
# 21:49:59 |
aneyman__ |
bhundven: the latter should be something contributors find out in their own testing |
# 21:50:12 |
aneyman__ |
before they even submit a pull request |
# 21:50:22 |
bhundven |
you'd be surprised |
# 21:50:28 |
bhundven |
:P |
# 21:50:46 |
aneyman__ |
by not testing it, you're substituting it with "let's not tell contributors they misspelled a variable" :P |
# 21:51:00 |
y_morin |
bhundven: I don't understand the "worker" issue. For Buildroot we have this: https://travis-ci.org/buildroot/buildroot-defconfig-testing/builds/136281896 <- notice: Total time more than 24 hrs |
# 21:51:24 |
y_morin |
bhundven: And it's not a paid account. |
# 21:51:26 |
bhundven |
hey, I mean if we want to run all the samples, by all means |
# 21:51:39 |
y_morin |
You may only get two builds in aprallel, but the number of builds is not limited. |
# 21:51:43 |
bhundven |
I don't *really* have a horse in that race |
# 21:51:49 |
bhundven |
I just ride the horse |
# 21:51:59 |
bhundven |
s/the/a/ |
# 21:52:35 |
y_morin |
bhundven: Anyway, just merge the first four commits of the series. I've already reviewed it. |
# 21:52:46 |
y_morin |
*them |
# 21:52:58 |
bhundven |
y_morin: and that is the other problem with github |
# 21:53:06 |
bhundven |
y_morin: the pull request is all of the commits |
# 21:53:11 |
y_morin |
bhundven: Yes, that is really annoying. |
# 21:53:38 |
y_morin |
bhundven: But don't come telling me you can't do a git fetch and a git cherry-pick locally! ;-) |
# 21:53:46 |
y_morin |
Hehe! ;-) |
# 21:53:48 |
bhundven |
I'm about a few more pain points from saying F#*K IT and going back to patchwork |
# 21:54:08 |
bhundven |
since patchwork has improved a lot |
# 21:54:28 |
bhundven |
and github is kind of useful, but super annoying |
# 21:54:32 |
lundmar |
my recent pull request just finished travis-ci (9 samples) in 50 minutes, not too bad for a free cloud service. |
# 21:54:50 |
bhundven |
lundmar: then add multilib and all of the other samples |
# 21:55:03 |
lundmar |
ehhh, make that 99 minutes. |
# 21:55:11 |
bhundven |
or more... |
# 21:55:13 |
aneyman__ |
hey, I am ok with dropping arm-multilib :) |
# 21:55:20 |
aneyman__ |
that was more of a proof-of-concept thing |
# 21:55:34 |
bhundven |
aneyman__: I have no problem keeping it, really. It *might* change over time. |
# 21:55:34 |
aneyman__ |
I know that sample alone builds for ~120 minutes :) |
# 21:55:39 |
lundmar |
30 samples would be about 300 minutes or so |
# 21:56:02 |
lundmar |
depends on the load ofc |
# 21:56:10 |
y_morin |
bhundven: Fifth patch is also OK to go. |
# 21:56:17 |
aneyman__ |
--with-multilib-list=aprofile has 28 multilib configurations :) |
# 21:57:18 |
y_morin |
bhundven: Sixth is OK too. |
# 21:57:36 |
y_morin |
bhundven: So for now: 1 through 6: you can cherry-pick! ;-) |
# 21:58:14 |
bhundven |
ok, I'm at work right now, and will be for a few more hours |
# 21:58:15 |
lundmar |
remember, github pull requesting is an optional feature ^^ |
# 21:58:42 |
y_morin |
bhundven: BTW, I'm speakign about commit ordering in the repository, not the list on github (in case that was not clear) |
# 21:58:56 |
y_morin |
bhundven: Oh, sure! |
# 21:58:56 |
bhundven |
lundmar: true, I can manually merge, but then I need aneyman__ to rebase and force push his work to update the PR with the rest of the changes |
# 21:59:10 |
y_morin |
bhundven: I tend to forget about timezones... |
# 21:59:33 |
y_morin |
bhundven: Yep, rebasing is often a pain-point... |
# 21:59:44 |
bhundven |
just keep posting stuff here, I'll look at the ctngbot logs later. |
# 21:59:51 |
aneyman__ |
y_morin: you can tell me how far you went, I'll create a separate branch with 1..X commits |
# 22:00:11 |
aneyman__ |
not a big deal, that's how I lived with multlib branch on top of fix-samples branch :) |
# 22:00:30 |
aneyman__ |
just tell me what the last reviewed commit is (i.e., the value of X) |
# 22:01:57 |
y_morin |
aneyman__: OK, I'll continue a little review, and when I stop, I'll tell both you and bhundven the value of X. |
# 22:15:50 |
y_morin |
aneyman__: OK, 1 through 10 are OK to go. Just minor nits that should not bar them from being applied, I guess. |
# 22:15:53 |
y_morin |
bhundven: ^^^ |
# 22:16:01 |
y_morin |
And now, sleep... |
# 22:16:12 |
y_morin |
quits : Quit: Nighty Night! |
# 22:23:15 |
aneyman__ |
bhundven: https://github.com/crosstool-ng/crosstool-ng/pull/403 |
# 22:23:28 |
aneyman__ |
running through travis-ci right now |
# 22:23:49 |
bhundven |
ok, I'm at work right now, and will be for a few more hours |
# 22:23:57 |
bhundven |
just keep posting stuff here, I'll look at the ctngbot logs later. |
# 22:23:59 |
bhundven |
:) |
# 22:24:38 |
aneyman__ |
ok |
# 22:25:02 |
aneyman__ |
mind if I'll start addressing Yann's comments on top of the existing changes? |
# 22:25:23 |
bhundven |
sure |
# 22:25:28 |
aneyman__ |
I am tired of rebasing them down to the commit that actually made that change |
# 22:25:51 |
aneyman__ |
especially in the cases where the corresponding code moved a few times or does not even exist in HEAD |
# 22:26:19 |
bhundven |
aneyman__: I am researching solutions to github, we may end up going to emailing PRs and using patchwork. I'm also not very happy with github's review process and management |
# 22:26:43 |
bhundven |
but as lundmar mentioned, there are some external review tools that work with github that I have not looked at yet |
# 22:27:11 |
bhundven |
but it is what we have now, painfully... |
# 22:27:58 |
aneyman__ |
I am fine with it either way - whether we stay with github or move to something else |
# 22:28:15 |
aneyman__ |
the biggest issue with GH is the disappearing comments during the rebase |
# 22:28:53 |
aneyman__ |
but it seems that did something to address that - is the "made a comment on an outdated diff" the indicator that the original comment commented upon was rebased? |
# 22:31:10 |
lundmar |
I think you will have a hard time finding something which is easier for devs to access than github. |
# 22:34:40 |
bhundven |
lundmar: easy to access doesn't always mean that it matches your workflow |
# 22:34:46 |
bhundven |
or improves it to that fact |
# 22:35:58 |
bhundven |
I know a few people that used to contribute to ct-ng stopped when we moved to github, because they didn't want to create a github account |
# 22:36:36 |
lundmar |
true, but still github is to git what fb is to social media. Just saying - if you want users to contribute github makes it easier for them. |
# 22:36:50 |
lundmar |
hehe, the opposite argument can be made |
# 22:36:56 |
bhundven |
indeed |
# 22:37:41 |
bhundven |
well, I need to bail so I can get work done so I can go home and review changes, or else I will have to go home and go back to work. |
# 22:37:48 |
bhundven |
quits : Quit: ttyl! |
# 22:51:31 |
diorcety |
quits : Read error: Connection reset by peer |