<beerman>
jaeger: will you make a commit in pkgutils when you feel like this is a safe/good option?
<jaeger>
sure, if we want to go this route. Needs more testing on my end. jue, any thoughts/objections?
<jaeger>
off to sleep for now
<beerman>
alright, good night
<beerman>
Romster: same question
<jue>
jaeger: no additional thoughts or ideas for that problem besides using tar instead of bsdtar if it works better, but no objections to add the --format options
<beerman>
tar would slightly complicate pkgmk code, imo, because bsdtar makes it easier to adjust which compression to use. Maybe I am wrong about gnu tar here..
<farkuhar>
it's not much more complicated to select a different compression using gnu tar. You can have the command include a pipe to the compressor (gzip, xz, ...) with gnu tar just writing to stdout.
<farkuhar>
and the compressor that receives the stdout is configurable by reading /etc/pkgmk.conf, as before.
<farkuhar>
last year when fakeroot was giving me headaches building java/ant/maven ports, for example, I wanted to use the --owner and --group flags of gnu tar (to avoid running pkgmk under fakeroot), and pkgmk was easily modified to do so: http://ix.io/3rJJ
<farkuhar>
of course, bsdtar has similar options already, but the exercise did demonstrate the viability of gnu tar for this purpose.
<stenur>
Or you ask upstream to get a binary modus that does not convert whatever the cost.
<stenur>
But really, libarchive only does iconv(3) from charset-of-archive to charset-of-locale.
<stenur>
(But its true my last sentence was wrong, regarding "if the foreign ball uses format=pax", if you simply grab+store what is in the archive.)
<jaeger>
jue: ok. I'm leaning towards leaving bsdtar in place but adding --format as a minimal change... but I will test it more on my systems over the weekend