| cpaelzer | xnox: should http://autopkgtest.ubuntu.com/packages/s/systemd/bionic/ppc64el be a force-badtest? | 07:33 |
|---|---|---|
| LocutusOfBorg | hello vorlon, do you feel ola syncing today? there is a versioned dep on python-protobuf left, but meh, since all the other delta is gone... maybe we can consider sync? | 07:39 |
| === ricab_ is now known as ricab | ||
| ginggs | tseliot: hi, do you have plans to backport the 430 driver to disco? | 10:25 |
| tseliot | ginggs: yes, to both disco, and to bionic | 10:26 |
| ginggs | tseliot: excellent! i've been experiencing lockups on a machine with 2 x TITAN X, upgrading to 430 seems to have solved it | 10:26 |
| tseliot | ginggs: I have it in a PPA here https://launchpad.net/~canonical-hwe-team/+archive/ubuntu/intermediate-kernel | 10:27 |
| ginggs | tseliot: thanks, i just backported the eoan version to disco in my PPA | 10:28 |
| tseliot | ok, well, it's the same | 10:28 |
| === ricab is now known as ricab|lunch | ||
| === ricab|lunch is now known as ricab | ||
| === ogra is now known as Guest59507 | ||
| juliank | jelmer: Is it expected that brz is about 2-3 times slower than bzr? | 15:47 |
| juliank | bzr info is 0.07s user, brz info is 0.25s user | 15:47 |
| juliank | um | 15:48 |
| jelmer | juliank: no, that's not expected | 15:48 |
| juliank | jelmer: Hmm, I should cProfile it | 15:49 |
| jelmer | juliank: if you can consistently reproduce it, we'd be very interested in a bug report (with profiling details would be awesome) | 15:50 |
| juliank | jelmer: https://bugs.launchpad.net/brz/+bug/1832868 | 15:57 |
| ubottu | Launchpad bug 1832868 in Breezy "2-3 times slower than bzr for simple things" [Undecided,New] | 15:57 |
| juliank | oh the filename does not end in callgrind, I'm bad at reading docs | 16:06 |
| vorlon | LocutusOfBorg: yep, sync of ola looks fine, done - thanks | 16:07 |
| juliank | callgrind stuff does not seem helpful anyway, it shows bzr using more ticks than breezy | 16:09 |
| juliank | despite being more than twice as fast in user time | 16:09 |
| * juliank is confused | 16:09 | |
| xnox | juliank: maybe try $ time bzr rocks | 16:28 |
| xnox | juliank: ditto brz ? | 16:29 |
| xnox | juliank: that should be "just startup time profiling of bzr" | 16:29 |
| juliank | xnox: 0.093 vs 0.303 | 16:34 |
| juliank | (real) | 16:34 |
| juliank | 0.072 vs 0.270 user time | 16:34 |
| xnox | horum | 16:35 |
| xnox | no idea if that is because brz has more plugins (shouldn't matter) or because it's just python3 | 16:35 |
| juliank | breezy seems to open a ton of .egg-info files | 16:37 |
| juliank | well read the PKG-INFO of each .egg-info | 16:37 |
| juliank | and the entry_points.txt | 16:37 |
| cjwatson | This sounds like the sort of thing that turns out to be due to importing pkg_resources somewhere | 16:42 |
| cjwatson | A while back I had a similar problem in LP and going around some lazr.* modules to avoid them importing pkg_resources unnecessarily was a big win, but it did involve some spelunking | 16:44 |
| cjwatson | Ah, so breezy.plugin imports pkg_resources and bzrlib.plugin doesn't | 16:45 |
| cjwatson | There's a faster version somewhere, let me see where I ran across it | 16:45 |
| cjwatson | importlib_resources | 16:45 |
| cjwatson | https://pypi.org/project/importlib_resources/ | 16:46 |
| cjwatson | Hm, but it doesn't have iter_entry_points | 16:47 |
| cjwatson | Anyway, it's definitely almost entirely due to that - "BRZ_PLUGIN_PATH=-entrypoints brz rocks" is close to the bzr time | 16:51 |
| === ogra is now known as Guest19708 | ||
| === heapall0c is now known as sm3lly | ||
| === sm3lly is now known as smelly_ | ||
| vorlon | juliank: it doesn't look like you upstreamed your changes from gnutls28 3.6.7-2ubuntu3 to Debian, am I missing it somewhere? | 20:21 |
| * juliank looks | 20:21 | |
| juliank | vorlon: Ah right, I upstreamed that to upstream git and did not follow up on it | 20:22 |
| vorlon | ah ok | 20:22 |
| juliank | https://gitlab.com/gnutls/gnutls/merge_requests/986 | 20:22 |
| juliank | But yes, I moved it to the test runner script, so it's probably good to upstream that part to Debian | 20:23 |
| juliank | Not sure if this needs to be upstream upstream | 20:23 |
| juliank | vorlon: Sent to Debian, no bug number yet | 20:29 |
| juliank | https://bugs.debian.org/930541 | 20:33 |
| ubottu | Error: debian bug 930541 not found | 20:33 |
| xnox | juliank: how can i make $ apt install openssl | 20:35 |
| xnox | juliank: upgrade everything that came from openssl source package, rather than just the /usr/bin/* binaries. | 20:35 |
| juliank | xnox: make openssl break the old versions? | 20:35 |
| xnox | juliank: sure, but like a generic apt syntax? $ apt install src:openssl => i.e. upgrade all binary packages that came from that source package? including like dbgsyms -doc -dev etc. | 20:36 |
| xnox | for any package =) | 20:36 |
| juliank | xnox: don't have syntax for that yet | 20:36 |
| xnox | juliank: it would be nice to do a wider match, when interactive, and have more exact syntax non-interactive, or something like that. | 20:37 |
| juliank | I think it will be something like ?upgradable(?source-name-exact(openssl)) | 20:37 |
| xnox | that would work too | 20:37 |
| juliank | xnox: I think apt should generally only upgrade entire source packages | 20:38 |
| xnox | i do wish for $ apt install src:openssl => like install source package unpacked into like /usr/src => to be used for Built-Using stuff | 20:38 |
| xnox | juliank: ooooh ! yeah, especially if it can without things getting broken, otherwise hold-back and upgrade only things it can. | 20:38 |
| juliank | Source package objects (to get all binaries of a source package) will be in apt 1.9.1 if all goes well | 20:39 |
| juliank | and patterns hopefully too | 20:40 |
| juliank | but I gotta write a parser and a lot of classes for them which will be annoying | 20:40 |
| juliank | apt install source packages I do recognize makes sense | 20:40 |
| juliank | Syntax-wise, I'd treat src as an arch and end up with openssl:src, though, I guess | 20:41 |
| juliank | but not 100% sure | 20:41 |
| juliank | maybe src:openssl is a name, and it's architecture is amd64 or something to have working cross-build dependency resolution | 20:41 |
| juliank | so that apt install src:openssl:amd64 installs openssl build dependencies for building amd64 packages | 20:43 |
| juliank | and src:openssl:ppc64el installs build dependencies for cross-compiling | 20:43 |
| juliank | xnox: So, I'm discussing this in #debian-apt, again | 20:49 |
| juliank | xnox: because I really want to have Build-Depends: src:pkg and apt install src:pkg | 20:49 |
| juliank | I think the latter is easy to do compared to the former | 20:50 |
| juliank | although, maybe not | 20:50 |
| juliank | xnox: Also, apt install src:openssl:src to recursively install source packages for all openssl build dependencies | 20:52 |
| juliank | :D | 20:52 |
| xnox | heh | 20:54 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!