[02:53] wgrant: have you looked at benchmarking pack.compression? [02:55] blr: No, but smart people have probably already set it sensibly. [02:55] There's nothing in our situation that should cause the defaults to be wrong, I don't think. [03:16] wgrant: can you think of any additional per repo defaults we need other than logallrefupdates and writeBitmaps? [03:16] wgrant: ensure_config is now pulling those values from git.config.yaml in turnip's root rather than defining them inline in the function [03:22] afaict there are no analagous per repo config variables for git repack's arguments [03:23] blr: some of them exist, but stuff like -l doesn't [03:24] http://git-scm.com/docs/git-config.html does warn that it isn't comprehensive, but I only see 'useDeltaBaseOffset', 'packKeptObjects' and 'writeBitmaps' [03:26] blr: git help config shows a number of pack.* options. [03:26] repack.* is just stuff that makes sense only for repack [03:30] wgrant: ah, I see --max-pack-size will honour pack.packSizeLimit if set [03:30] blr: Yeah. I don't think we care about pack size at the moment, though. [03:31] depth defaults to 50 [03:33] We probably want to set window and depth up a way, but also limit pack.windowMemory. I'd set windowMemory to like 2g for now, and we can tweak as we go. [03:34] hmm the repack docs don't mention if --window-memory takes a value from pack.windowMemory [03:35] if (!strcmp(k, "pack.windowmemory")) { [03:35] window_memory_limit = git_config_ulong(k, v); [03:35] return 0; [03:35] here I was thinking I could avoid looking at git source [03:35] } [03:35] --window-memory overrides that. [03:35] Heh, no. [03:36] It's pretty bad, but you get used to it. [03:39] wgrant: {depth: 100, window: 20}? (doubling defaults) [03:40] blr: Let's make both 250. That's semi-aggressive, and windowMemory will save us if it goes awry. [03:41] cool [03:41] An object to pass -f would also be useful, but it shouldn't be the default. [03:44] Actually, maybe we should make the default depth and window less aggressive, but allow it to be overridden. [03:44] Hmm. [03:44] 100/100 with options, perhaps. [03:45] 250/250 is slower than I'd remembered. [03:45] sure [04:07] wgrant: as we're now calling ensure_config on repo init, presumably it is no longer needed in PackBackendProtocol.requestReceived()? [04:08] however, existing repositories would be in an undefined state [04:09] given we don't have any means of changing configuration, ensure_config _for now_ can be treated idempotently, but something we may need to consider later [04:09] so I would be inclined to leave it there... [04:10] sorry, thinking out loud really [04:22] blr: The idea is that it's cheap and safe enough to run before every write operation. [04:23] So when we change it we don't have to do any bulk operation. [04:24] yep, that makes sense. [04:35] wgrant: repack-api ready for review. [04:35] blr: Thanks, looking. [04:37] wgrant: r158 has the git config yaml file, whoops. [04:53] wgrant: 3x gives you a stronger, more flavourful json extraction... [04:53] Heh [15:34] Hey! [15:34] I have a question about launchpad authorization tokens [15:35] I would like to authorize a script that would be running outside of my local computer with my credentials [15:36] Someone mentioned that when an app is authorized, there should be some credentials file created in ~/.launchpadlib which I should be able to re-use elsewhere [15:36] But I couldn't find it, but I didn't authorize for 'forever', so maybe that's why [15:43] sil2100: It lives in your desktop keyring, in fact. But maybe you actually want to re-login with Launchpad.login_with(credentials_file=...). [15:44] sil2100: Is the other computer entirely under your control? [15:45] cjwatson: not entirely, I was considering options on how to best set-up something like that on lillypilly or chinstrap [15:45] cjwatson: probably not safe enough [15:45] hmm [15:47] sil2100: You'd be better off with a separate bot user, or a little dedicated openstack instance or something [15:49] Indeed! [15:49] Anyway, thanks :)