| <<<Back 1 day (to 2018/04/17) | 20180418 |
moolc | tor8: (for the logs) https://www.boblycat.org/~malc/scratch/viper2.png | 07:53.21 |
| tor8: just to make sure, seen my for the logs message (apparently sent 15 minutes before you joined)? | 09:31.19 |
tor8 | moolc: I'm staying well out of the editor wars these days... | 09:33.44 |
moolc | tor8: well, good for your sanity... | 09:34.17 |
| tor8: fwiw, sebras here had only one problem building the best font viewer in the world (and said problem was corrected by http://repo.or.cz/llpp.git/commit/9f34d341c065684fb21ac0af1bb2e2083354075f) just saying.. you never know when your 1080p panel would die and you would be shopping for that mono-spaced unicorn | 09:36.16 |
sebras | I never understood the editor wars, I only want my peer to choose the one they are most productive in, I don't care which one. :) | 09:41.53 |
| s/peer/peers/ | 09:42.01 |
moolc | sebras: that's only because you are of insensitive Nordic who somehow avoided being infected by https://www.xkcd.com/386/ | 09:43.24 |
| ...dic type... | 09:43.34 |
sebras | moolc: perhaps. | 09:45.13 |
chrisl_laptop | tor8: You said the other day that the scripts for the third party mirror(s) were struggling - I assume that's mainly jbig2dec and it's the spltting out the of the git commits that's the issue? | 13:41.10 |
tor8 | yeah... the scripts need to run through the entire GS history to pull out the bits it needs | 13:41.34 |
| on every run | 13:41.36 |
| it's hugely inefficient and creates a *lot* of extra git objects and caches | 13:42.02 |
chrisl_laptop | The one for ijs, I've added that it does an MD5 on the subtree before and after pulling, and only does the rest if the hash differs | 13:42.38 |
| So it should only trawl the history is the files have actully changed | 13:43.06 |
| s/is/if | 13:43.16 |
| I figure the MD5 is a much cheaper operation | 13:44.35 |
tor8 | I have occasionally tweaked the script to start at a later point | 13:50.03 |
| but the jbig2dec stuff has some git grafts thrown like wrenches into the cogs. | 13:50.55 |
chrisl_laptop | It might be feasible to do something cheaper than the MD5 - like use the date | 13:52.06 |
Robin_Watts | tor8: How long does each run take? | 13:52.28 |
tor8 | Robin_Watts: a couple of minutes | 13:52.37 |
Robin_Watts | Then do we care? | 13:52.43 |
tor8 | I haven't cared enough yet | 13:52.51 |
Robin_Watts | If it was taking 6 hours a run, sure. | 13:52.55 |
tor8 | but I don't want us adding a half a dozen more subtree modules that need scripts and workarounds to get out | 13:53.13 |
Robin_Watts | My preference would be for lcms2mt to be generated similarly. | 13:53.15 |
tor8 | My preference would be to teach the rest of the staff to use git properly... :) | 13:53.49 |
chrisl_laptop | I was really hoping submodules would be more seamless by now.... | 13:54.12 |
tor8 | chrisl_laptop: they have improved significantly, but they're still not completely seamless | 13:54.36 |
| using relative submodule URLs has thrown a lot of people a curve ball; that was a bit unfortunate but required for mirroring on github | 13:55.18 |
Robin_Watts | I'm not sure that "learning to use git properly" entirely solves the problem. | 13:55.58 |
| Unless you're proposing that we move ghostscript over to using submodules for lcms2mt etc. | 13:56.17 |
tor8 | Robin_Watts: that is exactly what I was proposing. | 13:56.33 |
chrisl_laptop | Anyway, my point is even doing the recursive MD5, it cuts the time to negligible unless there are actual changes, so that is a solution | 13:57.29 |
Robin_Watts | I have to say that I kinda like the "gs is the golden repo, and all else is produced from it" solution. | 13:58.26 |
| It keeps stuff simple, and we know that as long as fixes have gone into that one place, they'll end up everywhere else. | 13:58.47 |
tor8 | Robin_Watts: as long as I don't have to maintain it anymore :) | 13:58.50 |
| chrisl_laptop: right. that could be a useful improvement. feel free to take out my scripts on casper and rework them. | 13:59.43 |
| in /home/tor/bin/mirror-jbig2dec.sh | 14:00.08 |
| and mirror-thirdparty.sh which calls the extraction scripts and does the final pushing to the master repos | 14:00.52 |
chrisl_laptop | tor8: I'm going to have play with using the date instead of the MD5 - once I know which works okay, I'll rejig your scripts | 14:01.24 |
Robin_Watts | chrisl_laptop: So you md5 each *file*, or the entire tree ? | 14:01.31 |
| Isn't there a git command that gives you the SHA for each file (cos it's got that precalculated) | 14:01.57 |
| so that would cut the calculation time down a lot. | 14:02.10 |
chrisl_laptop | Robin_Watts: I MD5 each file, then MD5 the results of the MD5s | 14:02.20 |
Robin_Watts | Right, so you could MD5 the SHAs instead. | 14:02.35 |
tor8 | chrisl_laptop: git cat-file -p $(git cat-file -p HEAD | grep tree | awk '{print $2}') | grep jbig2dec | 14:03.03 |
chrisl_laptop | I didn't think git held SHAs for individual files, I thought only for the entire repo | 14:03.03 |
Robin_Watts | git ls-files -s filename | 14:03.05 |
| chrisl_laptop: For every file, I believe. | 14:03.15 |
chrisl_laptop | OKay, that's easy enough to use | 14:03.42 |
tor8 | that should get you the recursive sha1 for the jbig2dec subdirectory in the current git HEAD | 14:03.44 |
Robin_Watts | git ls-files -s jbig2dec | cut -something | 14:04.02 |
chrisl_laptop | I'll have a mess with it next time I have a little time | 14:04.47 |
Robin_Watts | git ls-files -s jbig2dec | cut -d" " -f 2 | 14:04.53 |
| or: git ls-files -s jbig2dec | cut -f 1 if you want to include permissions and SHA. | 14:06.26 |
| or: git ls-files -s jbig2dec --name-only <revision> | cut -f 1 apparently. | 14:08.47 |
chrisl_laptop | I have to take my car for an MOT (it's first!) - not sure if there's wifi at the garage..... | 14:10.55 |
moolc | is URW++'s Z003 - ZapfInno alternative? | 17:13.05 |
jmg | hey all | 23:14.23 |
| looking for some people that know kobos well, maybe the koreader guys? | 23:14.38 |
| Forward 1 day (to 2018/04/19)>>> | |