marlin has quit [Quit: Ping timeout (120 seconds)]
marlin has joined #crux
goodlings8 has joined #crux
marlin has quit [Quit: Client closed]
pfr has joined #crux
pfr has quit [Quit: pfr]
pfr has joined #crux
pfr has quit [Client Quit]
pfr has joined #crux
pfr has quit [Client Quit]
iNomad has quit [Ping timeout: 240 seconds]
iNomad has joined #crux
iNomad has quit [Ping timeout: 268 seconds]
ppetrov^ has joined #crux
<cruxbot>
[opt/3.7]: prt-utils: updated templates for prtcreate
<ppetrov^>
thanksd for the updates, farkuhar
<ppetrov^>
where does the cd ${name#r4-} comes from in, prtcreate r?
<farkuhar>
ppetrov^: it comes because you named all your R ports starting with "r4-". On the assumption that others would follow your example, the bash parameter expansion strips off the leading "r4-", which often yields the correct directory name (modulo capitalization).
<farkuhar>
ppetrov^: if your wish for an expanded contrib repo was not in jest, here's a sobering thought. Adding all the dependencies for jupyterlab would expand the contrib repo by about 50 ports. I'd prefer a more easily-navigated portdb, which would involve someone following your example with r4-modules, and hosting a dedicated "scipy" repo.
<farkuhar>
Putting together a "scipy" repo should be straightforward: just install a python3 virtualenv, and run `pip3 install jupyterlab` in that venv to see which python packages would be pulled in as dependencies. Then write CRUX ports for anything that isn't yet in the portdb.
lavaball has joined #crux
<ppetrov^>
farkuhar, name=r4-foo is fine, however to build the package, it should be "cd Foo"
<ppetrov^>
sometimes they have a capital letter
<ppetrov^>
anyway, it's a cosmetic thing and it's perfectly fine like this. Also, i seriously doubt anyone else will be interested to package R modules for CRUX (which are in fact "R libraries", but I wanted the ports collection to be called analogously to the perl modules)
<farkuhar>
sepen's first response in that thread gives an indication of the guideline they tried to follow back then: if you list any dependencies outside core opt xorg, you should provide them in your repo. Nowadays the practice is more relaxed, and nobody complained that the r4-modules ports all depend on contrib/r.
<farkuhar>
but contrib/r is just _one_ port outside core opt xorg, whereas the hypothetical "scipy" repo I outlined above would involve many more dependencies currently residing in contrib.
<farkuhar>
following the guideline that prevailed in 2008 would require that a scipy repo include dups of all those dependencies currently in contrib, not just the ~50 additional dependencies of jupyterlab that don't yet have CRUX ports written for them.
<ppetrov^>
farkuhar, ok, i get it
<ppetrov^>
thanks for the clarification. However, a bunch of my ports (in ppetrov) depend on stuff from contrib, so that, too is not following the recommendations closely
<ppetrov^>
even the mate ports depend on stuff from contrib, or I am mistaken?
<farkuhar>
ppetrov^: short version is, it's now considered "cleaner" to avoid duplicate ports than to follow an ancient recommendation carried over from the early years of CRUX.
<ppetrov^>
right
<ppetrov^>
so, anything i should fix/improve in my stuff?
<farkuhar>
no, you've done a great job! Hopefully someone will be inspired by your example and create a "scipy" repo as I outlined above, and then we won't have zorz dropping uninformed insults like "all Linux distros except for CRUX have a package for pandas".
<ppetrov^>
huehuehue
<ppetrov^>
and thanks for the kind words!
<ppetrov^>
i guess when R becomes major version 5, i will have to change to r5-modules
<ppetrov^>
and I have missed what zorz had complained about
<ppetrov^>
i think Arch had some tool for automatic creation of PKGBUILDs for pip or sth, but it had major issues with resolving dependencies
ln43 has joined #crux
ln43 has quit [Client Quit]
<farkuhar>
ppetrov^: we already had cpan2crux, and thanks to your efforts we have cran2crux. Now somebody just has to write pypi2crux, and we can generate native CRUX ports for the entire python package index.
SiFuh_ has quit [Remote host closed the connection]
<ppetrov^>
cpan2crux should be in contrib, imho
<ppetrov^>
if i switch to python for some future project, i might try to do a pypi2crux, but don't hold your breath
<ppetrov^>
if i understood correctly, there should not be /usr/share/doc, right? These from contrib install stuff there:
<ppetrov^>
contrib/enchant
<ppetrov^>
contrib/hwloc
<ppetrov^>
contrib/mpv <-- has mpv.conf in /usr/share/doc/mpv/
<ppetrov^>
while these have generated documentation there on my machine:
<ppetrov^>
contrib/lv2
<ppetrov^>
contrib/sord
<ppetrov^>
files are not the the ports' .footprint
tarxvfz has joined #crux
<farkuhar>
man 5 Pkgfile says "... Apart from these global rules, the definition of junk is often a matter of personal taste." Online documentation in the spirit of GNU info pages is frowned upon, but example configs and concise reference files are usually acceptable.
<ppetrov^>
all right
lavaball has quit [Remote host closed the connection]