fishe has quit [Quit: Connection closed for inactivity]
braewoods has quit [Quit: WeeChat 2.8]
braewoods has joined #crux-devel
braewoods has quit [Quit: WeeChat 2.8]
braewoods has joined #crux-devel
<dlcusa>
farkuhar, I like pkgmeek after a quick glance, but am unable to exercise it for now.
<farkuhar>
dlcusa: thanks for the feedback. I'm following the dictum "publish early and publish often", so what you see is still a WIP.
<farkuhar>
You can follow the git log to see how much it's been revised over the past 24 hours.
<farkuhar>
I was in the middle of a debugging session when my SSH connection to the crux machine broke down. There might be a vim swap file to recover when I finally do manage to log back in.
<farkuhar>
dlcusa: I just pushed more bugfixes to the git host. This latest version is passing more unit tests (at least on the handful of ports that I throw at it).
groovy2shoes has quit [Remote host closed the connection]
groovy2shoes has joined #crux-devel
<beerman>
jaeger: did you build any newer test iso yet? do you see any roadblocks for 3.7?
<beerman>
farkuhar: sorry i had no time to test your script, and i am unsure if I should take the time to do it. I fear that chances are slim to just abandon pkgmk and jump on something else, so even if I would like it, I wouldn't be able to use it in a container to build ports either. anyway, cool cool.
<farkuhar>
beerman: no problem, I was still ironing out the last several bugs. I just pushed the latest set of fixes to the git host.
<farkuhar>
As advertised at the top of the file, it's really just the same code as the original pkgmk, formatted so that you can see more of the underlying algorithm on one screen. It might look like a wall of text at first glance, but after chunking the code into manageable units, you can see immediate benefits for maintainability.
<farkuhar>
it's now down to 561 LOC, and 50 of those lines are comments. Even at such a small size, it still has support for tarball renaming (and experimental support for parsing git sources).
<farkuhar>
oh, and i think it addresses both FS#1851 (Pkgfile sourced before signature verified) and FS#1763
<farkuhar>
i know you closed FS#1763 as a "wontfix", but the issue was real: if you grep for "$PKGMK_SIGNIFY -q -C -p", you'll see a line that trusts the signature file itself to provide all the filenames. It should have looked at the ${source[@]} array instead and explicitly listed all the files needing verification.
<jaeger>
I ran into some issues with my build VM that I've fixed but haven't yet built another ISO. I will make a point to do that in the next couple days, been crazy busy
<beerman>
jaeger: cool, would be appreciated. i might be able to have hardware to do a fresh install
<beerman>
farkuhar: if there is interest from other core maintainers to adopt this in favor of a more cleaned up pkgmk my interest might change, but as is right now, I got my hands busy with life and taking care of other peoples ports among my own ;)
<beerman>
if you look closely at 1763 - the reason for closing was more precisely "due to inactivity". in general I would like to adopt some sort policy to close issues more quickly. i dont like seeing them hanging around, but i have made that clear on numerous occasions already and there is just this much effort i will put into it
<farkuhar>
beerman: I share your sentiments about FlySpray tickets that linger around so long with no response. It's been more than a year since #1851 received any comments, and i think part of the reason is that the existing pkgmk has too much repetition of code, making it harder to maintain. my intent with pkgmeek is to make life easier for future maintainers :-)
<beerman>
sure, i do appreciate it. win over jaeger and jue and you certainly get more of my time for it too.
<beerman>
or redo it all in rust, i'd be all over that no matter what ;)
<farkuhar>
as for #1763, when the ticket creator failed to follow up on our requests for more evidence of a problem, you made the right call to close it. But it did inspire me to take a fresh look at the pkgmk code and realize again how much cruft it had accumulated over the years.
<farkuhar>
I'm sure that rust has a lot of advantages over shell scripts. One shortcoming of bash that just bit me today is the lack of support for more than one level of dereferencing. I was trying to save keystrokes by asking it to expand ${PKGMK_$VAR} as $VAR took on all the values in a list, but bash refused to cooperate.
<farkuhar>
i guess i was spoiled by all my experience coding in Perl last week. Returning to bash to write pkgmeek was hardly a frictionless move.
<braewoods>
farkuhar: you need a hack involving eval. something like: eval echo \${PKGMK_$VAR}
<groovy2shoes>
farkuhar, or: v=PKGMK_$VAR; ${!v}
<groovy2shoes>
(the manpage makes it sound like ${!PKGMK_$VAR} ought to work, but i can't get bash to accept that for the life of me. it works with the extra assignment, tho)
<farkuhar>
braewoods, groovy2shoes: thanks. i was only looking to solve that problem during a print debugging session (not in the code that gets shipped), but i'll make a note of those hacks for future reference.