Samba cleanups

So, it seems like there’s a general convergence in the ether, because for the past few weeks I’ve been sending emails back and forth to our samba team, about the messy samba ebuilds in portage. Meanwhile, both slarti and beu noticed the messiness and talked about it.

So, trying to shuffle that in between work-work and gentoo-work took a while. Anyway, the result was the new ebuild that I put into portage yesterday. Major props to beu for testing it for me, between cvs commits and rsync updates. It looks like the new ebuild is about half the size of the old ones, which is a plus. The cleaning, however, has only just begun. The next step is to move the other two ebuilds to use patch tarballs offada mirrors as well. The final step, before I finally relinquish the packages back to our samba guys will be to move the configs outtada files/ directory and ontoda mirrors.

The topic, the topic, what topic?

See, here’s the thing Donnie, I think the planet should honestly make available all the blogs from a dev who chooses to be on it. I think it’s a silly rule to have it be only on-topic. What is on-topic really? I mean, just because you don’t say (in your case, xorg) in your blog entry, does not mean that what you talk about (even if it is about Constantine getting kicked off) affects your work as a dev. Just as a by the way, constantine’s kicked-offness doesn’t affect me, personally, but I wanted to see how much off-topic I could get with this post. so ha! Anyway, I do think it’s a bit silly to have to go to three different sites to read your blog, instead of one-stop shopping. Kulleen out.

Xterm and Ryan Seacrest

So, I made a HUGE mistake the other day when I put xterm-201 into portage. First of all, I hate when people do the straight to stable thing on architectures they know nothing about. So, that means I hate myself this time around.

The other thing is — well, now that xterm has a toolbar, we got requests to be able to enable it. I used to do it off the local USE flag. I decided, instead to just enable its building, but suppress it from automatically showing up when you fire up an xterm. So, I made a little patch, and released into the wild, after having tested it on my box.

Then, I tested some more on my box, and it turns out that a command like this:
xterm -sl 14000 -sb -hold -geometry 150×10+10-30 -title “Camera” -e “time sudo nice -n -10 sudo -u $USER python $*” &

doesn’t actually work. In fact, it doesn’t work at all. Well, the xterm shows up, but it’s all blank. It’s hard to see whether the command in it is actually running, because the xterm is blank, and non-responsive. And trying to terminate those programs is a pain. So, like, I dunno what to do about this — but if someone out there knows, please let me know. I’m rather tempted to hard mask it in 24 hours without input, to be honest. (Though, I am planning on taking my problem to the excellent and brilliant Thomas Dickey and maybe he can guide me through this jungle.

In other news, can you believe Constantine got kicked off? Shocking, when people expected otherwise.
Not that I’m a big fan, or any fan of the guy, but who’s voting for the other two yahoos?

Even More -messy -foptimizations

So, now there’s bug 90663 which shows a user clearly using compilation flags gleaned from some 1337 ch4nn31 on some freaky irc network somewhere. People, if you’re going to be filing bugs, please at least tone down your cflags before you do. Bug reports like this are a waste of:

  • Gentoo developers’ time
  • Your own time

It’s a small enough list that you can probably memorise it. Please pay close attention to the first item there.

Happy Birthday, Sindhuja

Yeah, that’s my baby sister, she turns 25 today.

So, bunch of things. The whole co-habiting of kerberi is abandoned. There was a stunning lack of feedback as to motivation and method, so they’ll remain blocking. However, they will be cleaned up. I will make sure to at least make them both non-clobbering against e2fsutils & co.

Also, if you’re going to use crazy-assed compiler flags, the very least you can bother doing is reading the documentation about them. Honestly, people, gcc provides great man pages and info pages, and the website is chock-full of information about each flag. Don’t just listen to some l337 dud3 in some hokey irc channel on some hokey network — chances are that’s not freenode/#gentoo, but who knows.

And while you’re changing your cflags to put some cool looking flags in there, for the sake of all that is holy, please don’t screw with other settings that you know absolutely nothing about!!

It’s cases like these that are the cause of ridicule to Gentoo itself, to be honest. When you read /. posts or comments to linux news stories, or go on over to funroll-loops, it’s you they’re talking about. But then they generalise to the rest of us. And if you file a bug about your broken cflags set up breaking gentoo packages — if I happen to wrangle it, expect to see:

gcc -fuck-off-with-your-flags thisbug > closed

as your comment.

dnsmasq is the bee’s knees

Donnie, I’ve been using dnsmasq for a year now, ever since Ciaran mentioned it to me in #gentoo-sparc and it has taken its place in my top 5 favourite pieces of software. I don’t know that I would ever choose to use anything else. Granted, I don’t have a need for it in any large enterprise so I don’t know how well it would scale. But for smallish networks, the thing kicks ass.

And, by the way, are the writers of 24 all addicted to crack? Seems like the last month or so has been full of fillers, rather than any real plot development. I mean how is it the biggest twist you never saw coming when you saw it coming for the 3 episodes prior? And please, stop with the torture already.


OK, so here’s an update to the whole heimdal/mit crisis. I sent an email to gentoo-dev mailing list soliciting opinions on two options, which I shall outline below. There is no link to the email because the mailq on the gentoo mail server is apparently backed up. I’ll update this entry with the link when there is one.

In the meantime, here’s the two options:

1. Continue on as I have been, with prefixing things with mit- and heimdal- and then making symlinks.
2. Install mit and heimdal into /usr/libexec/kerberos/[mit,heimdal] without renames and making symlinks.

So you see symlinks get made either way. I think the libexec route (or /usr/lib/misc, as SpanKY favours) is the easier of the two. The other concern which Uberlord finally vocalised in #gentoo-dev is: is all of this worth it? It wouldn’t be possible, even if we removed the blockers on each other, that some package foo would compile against mit, but work against heimdal. So, then, what is the point? I don’t know. But if you do, please let me know.

Heimdal update not in portage

OK, so as promised I started working on heimdal’s new ebuilds for the whole relocation-no-stomping routine that you’ve been doubtless on the edge of your seat about. And I hit a snag. Not a snag, as much as an annoyance. The autotools provide this really nice configure flag that you can see in use in mit-krb5 (whose use karltk taught me for making sylpheed-claws co-exist with sylpheed back in the day when I actually maintained stuff). This would be the –program-prefix flag. Of course, for -claws, we used –program-suffix. So, for mit-krb5-1.4, this causes mainly the clients to be installed as mit-$clientname. So the telnet and su and all that crap is just mit-telnet and mit-su and what have you.

Heimdal, just to make my life complicated, goes a step further with –program-prefix thing. They rename _every_ binary. And every manpage. But of course, the symlinks that the makefile installs (there’s a symlink for kadmin to ktutil I think that doesn’t point to heimdal-ktutil, but to some non-existent ktutil; there’s also a whole crapload of symlinks in the man directories, but they also point to non-existent manpages). So now I have whack the configure/make scripts to correct this and also report this situation upstream. Stay tuned.

Retiring your developer cloak

What is it with Gentoo Developers leaving lately? Two people, scandium and chriswhite, have left in the last two weeks. Ordinarily, I’m used to the turn over in Gentoo development, but honestly not being on the -core list since January (until my dev boxes arrive here, don’t get alarmed), I miss out on the news and find out in funny ways. I guess that not being on devrel lists plays into that. Maybe it’s just my ego, but I wish I’d known their intentions before I found out and before they left so that, I dunno, I might have at least had the chance to change their minds.

And “why?” you might ask. Well, it’s very simple, the two of them were great developers. ChrisWhite took so much on in the media-video category (primarily, but he wasn’t afraid of media-* and lisp and fortran, for crying out loud) and fixed it up nicely. And that was just to start with. He handled the relationship with Real (a relationship that is NOT easy to handle, trust me). He handled various other packages. In other words, a dependable developer.

Scandium was responsible for all things cvs in portage. No, not the *-cvs packages that the ricers seem to love, but the actual dev-util/cvs stuff. You’ll find that he handled other stuff in the dev- categories like pike, a language I’ve certainly never heard of.

Anyway, as of this writing, I have no idea why either of them left, and I’m a little pissed off about it, because at this stage, it’s probably way too late to change either of their minds. Hey, I can try.

Anyway, since I’m not on -core I couldn’t give them a send off, so I suppose this qualifies, even though it’s much more public.

mit-krb5 is settled

OK, so here’s a nice update. I finally settled the installation scheme for mit-krb5, effective from 1.4 onwards. I’m working on the kerberos-update script (not kerberos-config anymore) to install proper symlinks, but that won’t be finished till the weekend.

Does anyone here actually use krb4 themselves? If so, what stuff that needs krb4 needs a hardcoded path into /usr/athena? I looked at kth-krb (because in all fairness kerberos-update should actually be able to work with krb5 and krb4), or I could pretend krb4 doesn’t exist and call it krb5-update. What do you think?

So updated todo:

  • Fix up heimdal — install two snapshots, one 0.6 and one 0.7(?)
  • Finish this kerberos-update (for now, it’s kerberos-update not krb5-update) script
  • Pray

PLEASE PLEASE comment here about the krb4 questions.