<SiFuh>
zorz: farkuhar needs to include git into prt/pkg utils
<zorz>
i am not fun of git in prt/pkg
<zorz>
arch now uses git... but i always have in my mind the xz backdoor
<SiFuh>
zorz: Don't care if you have fun or not
<zorz>
ahhaha
<zorz>
fan
<SiFuh>
Wife just tried to have a go at me and lost.... Hahahaha
<zorz>
better with tags releases... anyway instead of git, can always download the snapshot.... so what all the husle about ?
zorz has quit [Quit: leaving]
SiFuh has quit [Remote host closed the connection]
SiFuh has joined #opencrux
farkuhar has joined #opencrux
<SiFuh>
farkuhar: farkuhar is here
<farkuhar>
zorz: you make a good point about the possibility of downloading git snapshots with the existing pkgmk. The main advantage of native support for the git protocol is that incremental updates become much faster, since you only need to download the changeset and not the entire source tree.
<farkuhar>
As for the feature to restart a failed build from an already-unpacked source tree, Alan gave us a patch for that behaviour back in 2016. I'm too lazy to dig up the mailing list post right now, but here's the overlay port in his repo: https://www.mizrahi.com.ve/crux/ports/pkgutils/
<SiFuh>
and don't forget to create a makedepends
<farkuhar>
As I see it, the distinction between makedepends and runtime dependencies is more critical if we were planning to release a binary distro. For a source-based distro, it's arguably less important to make such a distinction.
<SiFuh>
farkuhar: Why install something you only used once for compilation?
<farkuhar>
SiFuh: "something you only used once" <- That might be an accurate description if you never upgrade the package. But when you want to build a newer version, then having the "makedepends" already installed should be faster than rebuilding them on a just-in-time basis.
<farkuhar>
The conversation in #crux is inspiring me to re-read the curl man-page to learn what the --retry 3 flag is supposed to do, and why the upgrade 8.14.0 -> 8.14.1 seems to give different exit statuses than before, when this flag is passed.
<farkuhar>
Hmm, /usr/share/man/man1/curl.1.gz is rather inconsistent in its escaping of hyphens. For most values of MANPAGER, I get a rendered man-page where some of the hyphens are visible and others are invisible.
<farkuhar>
I haven't played around with every conceivable value of MANPAGER, though. Maybe there are some programs that are more forgiving of mistakes in roff syntax, and they'll render the hyphen even if the man-page author forgot to escape it properly.
<SiFuh>
farkuhar: But it is already rebuilt. In is in an *#*gz pkg.
<SiFuh>
You'd build the package, use it for make depends and install what you made but the built makedepends pkg would still be a built archive on the system meaning you won't have to build it again unless it is updated
<farkuhar>
SiFuh: Yes, the makedepends might already exist as built packages on disk. And arguably it would take up less space in the long run, to extract them into a temporary chroot or work directory only for the purposes of building the dependent package. But this attempt to micromanage the disk utilization is starting to veer away from the KISS principle, imho.
<farkuhar>
I can see the rationale for separating runtime dependencies from buildtime dependencies, if we were to make binary package management a first-class citizen. Certainly, users of pkg-get would appreciate the disk savings from not having to install a bunch of compilation tools. But if building from source is the default, then keeping that pathway free from unnecessary complexity should be a more relevant
<farkuhar>
principle.
<SiFuh>
And how is our suggestion an unnecessary complexity if it is built into the pkg/prt utils
<farkuhar>
The distinction between runtime dependencies and build-time dependencies is *not* built into the current pkg/prt utils; it's a feature that would have to be added. Every time an additional feature is proposed, it's worthwhile to ask whether that feature introduces more complexity than can be justified by its benefits.
<SiFuh>
farkuhar: You trying to avoid it, making excuses to get out of it?
<farkuhar>
When darfo chimed in to the discussion of a bloated core repo several years ago, he said he appreciated that CRUX never went the way that Arch did, introducing into their PKGBUILD one innovation after another. Even if the core repo got bloated with ports like linux-pam and dumb_runtime_dir, the Pkgfile format itself remained almost as simple as when Per first designed it.
<SiFuh>
What's that go to do with the price of tyres?
<farkuhar>
See also Anton's post to the crux-devel mailing list in response to Heidi's "use flags" proposal (2006-08-29).
<SiFuh>
looking up posts is an unnecessary complexity
<farkuhar>
Looking up the price of tyres is an unnecessary complexity in a discussion about pkg/prt utils.
<SiFuh>
Well it is decided. zorz already agreed. You are implementing it
<farkuhar>
zorz was writing Pkgfiles that invoked `prt-get install` within the build function itself, until ukky set him straight. I wouldn't trust zorz to be the best judge of what constitutes unnecessary complexity in the Pkgfile format.
<SiFuh>
I think ye be lazy ;-)
<farkuhar>
I'd like to hear what dlcusa, emmett1 and ukky think about the proposal to have a separate header for makedepends. So far the only votes definitively in favor of this proposal are from zorz and SiFuh.
<SiFuh>
I've been busy, but I have been taking a couple of minutes out of the day to work on setup. Slowly..... going forward as time goes by
<farkuhar>
The curl man-page has 3 instances of features added in version 8.14.*, but none of them relate to the exit status when --retry 3 is passed. 7.66.0 and 7.71.0 are the only versions cited in connection with new --retry-* behaviour.
<farkuhar>
jue himself was responsible for reporting #17545 on the curl github instance, for which the ports tree now has a patch. The errors being reported in #crux are not fixed by this patch, but instead by downgrading to 8.13 as beerman did.
<SiFuh>
did beerman block jue?
<farkuhar>
No reason for beerman to block jue. The latest misbehaviour of curl was only reported upstream within the last 5 minutes: https://github.com/curl/curl/issues/17554
<SiFuh>
So he will block jue soon then? Because beerman hates being told ports are broken and thinks of people as useless whinges
<farkuhar>
All the recent updates to curl were pushed by jue, not beerman. So why would beerman take offense at a report that curl 8.14.* now gives different exit status when --retry 3 is passed?
<SiFuh>
because he is a ninny
<farkuhar>
In the meantime, `prt-get lock curl` might be necessary to prevent the sysup operation from upgrading curl to the 8.14.* version that breaks pkgmk downloads.
<SiFuh>
Seems to be an unessessary complexiity
<farkuhar>
If the different exit status was intentional, then I would expect it to have been documented in the curl man-page. The fact that the man-page mentions other features introduced with 8.14.*, but not the change in exit status, suggests that it was a bug that crept in by accident.
<SiFuh>
You expect too much.
<SiFuh>
farkuhar: That latest episode of Joe Rogan with Oz was pretty insane
zorz has joined #opencrux
<zorz>
la la lo lo la la la
<zorz>
where is this gay SiFuh_ :P
<zorz>
oooo farkuhar is online :P
<SiFuh>
zorz: farkuhar told you to fuck off witho your stupid makedeps
<SiFuh>
He said you are as dumb as a plank of wood and that all source based distros should be bloated with software for compiling even if you don't need it anymore
<SiFuh>
Said you are an unessessary complexiity
<zorz>
seriously :P
<zorz>
well done then :P
<SiFuh>
No, but he tried. I just paraphrased it better
<SiFuh>
[22 45 40] [farkuhar> I'd like to hear what dlcusa, emmett1 and ukky think about the proposal to have a separate header for makedepends. So far the only votes definitively in favor of this proposal are from zorz and SiFuh.
<zorz>
i did not vote :P
<zorz>
ahahahah
<zorz>
SiFuh: now i am reading the log... is like reading a novel :P
<zorz>
sifuh i did sbase, if i add ubase plus runit... no more shadow, linux pam.
<SiFuh>
zorz: Yes farkuhar has a way to turn IRC logs into the Mahabharata
<zorz>
let me try it
<SiFuh>
Yeah I saw ubase, and preferred sbase
<zorz>
you had the idea... and as always i said its for you(plural) to decide.
<SiFuh>
chinarulezzz hated the original system and uused mostly suckless tools. sinit, sbase and so on.
<zorz>
its better.... let me see how can i get getty from ubase... so i can have runit working and a system to work on
<SiFuh>
Well, we don't have to stick with the actual packages CRUX had originally. They were only used back then because there wasn't much of a choice.
<zorz>
but decide what you like for # vim: filetype=sh
<SiFuh>
No tab vs spaces is up to everyone to decide. I use 2 spaces. But if decided, I can shift to 4 spaces or tabs. I just think there needs to be a consistant format style.
<zorz>
i have the default of vim/nvim
<SiFuh>
Also I personally line up the \ with all others just to make a nice box like dlcusa does :-P
<ukky>
I prefer "$PKG" except as passing multiple space-separated parameters; and I never use aligned '\'. Indentation using tabs instead of spaces for both C and shell scripts.
<SiFuh>
ukky: I can accept that. But yes, should at least have " ", Tab vs spaces? Don't care. I choose what everyone chooses. The \ thing is just a nice block.
<ukky>
And we should never use pre-generated ./configure from upstream if package uses autotools.
<SiFuh>
ukky: In my core ports, only one package requires autotools.
<SiFuh>
Hence the bloody, I don't want to install autotools for one freaking port debate above about makedepends
<ukky>
zorz: your Pkgfile looks good. And kudos for using autoreconf.
<zorz>
ukky grande master :P
<zorz>
anyway... iam going to have sex with ubase
<zorz>
cheers!
<ukky>
zorz: I am just a bystander, with very firm opinion for how SW should be build and developed.
<zorz>
echo ">> NOTE: Cyrus SASL not found – Mutt built without SMTP support."
<ukky>
zorz: mutt should not be compiled with SMTP support. mutt is MUA and should not know details on how to retrieve and deliver mail. That's Unix way, a tool for every small task.
<zorz>
yes
<zorz>
so add an option... if cyrus installed then + smtp
<zorz>
okay i will fix it later
<ukky>
zorz: you do not have to change your Pkgfile, if in your opinion mutt should use SMTP.
<zorz>
no its better not to have the smtp option if cyrus is not installed
<zorz>
if cyrus installed then add the option.
<ukky>
zorz: NetBSD builds mutt with these: --enable-external-dotlock --enable-pop --enable-imap