that’s whacked

vimperator

Thursday, Jan 17. 2008  –  Category: OpenSolaris

best. firefox extension. ever.

seriously. i’ve been using this all day and love it. between Witch, Quicksilver, and now Vimperator, my hands seldom need to touch a mouse now.

if only i could find a keyboard way to move a window without having to use the mouse. oh well.

sb on os @ shdh, giving up for the night

Sunday, Jan 6. 2008  –  Category: OpenSolaris, Songbird

well. i got to the point where it segfaults now. 11:45pm. i think i’m calling it a night.

[stevel@grommit:dist] 554$ ./songbird
Segmentation Fault (core dumped)
[stevel@grommit:dist] 555$ mdb core
Loading modules: [ libc.so.1 libuutil.so.1 ld.so.1 ]
> $c
libc.so.1`_lwp_kill+0x15(1, b)
libc.so.1`raise+0x22(b)
libxul.so`__1cNnsProfileLockSFatalSignalHandler6Fi_v_+0xe6(b, 0, 80461c4)
libc.so.1`__sighndlr+0xf(b, 0, 80461c4, fceaf2cc)
libc.so.1`call_user_handler+0x2b8(b, 0, 80461c4)
libc.so.1`sigacthandler+0xc2(b, 0, 80461c4)
libmozjs.so`JS_BeginRequest+0x34(81543d8)
libxul.so`__1cXnsScriptSecurityManagerEInit6M_I_+0x7e(8162c78)
libxul.so`__1cXnsScriptSecurityManagerYGetScriptSecurityManager6F_p0_+0x57(
8046568, 81442a8, fdecc4d0, 0, 8046568, 81116bc)
libxul.so`__1cbIConstruct_nsIScriptSecurityManager6FpnLnsISupports_rknEnsID_ppv_
I_+0x33(0, fe2f29f0, 8046568)
libxul.so`__1cQnsGenericFactoryOCreateInstance6MpnLnsISupports_rknEnsID_ppv_I_+0
x1e(81628d0, 0, fe2f29f0, 8046568)
libxul.so`__1cWnsComponentManagerImplbACreateInstanceByContractID6MpkcpnLnsISupp
orts_rknEnsID_ppv_I_+0x121(8111678, fdd7c9e4, 0, fe2f29f0, 8046568)
libxul.so`__1cWnsComponentManagerImplWGetServiceByContractID6MpkcrknEnsID_ppv_I_
+0xfb(8111678, fdd7c9e4, fe2f29f0, 80465ec)
__1cOCallGetService6FpkcrknEnsID_ppv_I_+0x52(fdd7c9e4, fe2f29f0, 80465ec)
__1cYnsGetServiceByContractID2f6kMrknEnsID_ppv_I_+0x25(80465fc, fe2f29f0, 
80465ec)
__1cNnsCOMPtr_baseZassign_from_gs_contractid6MknYnsGetServiceByContractID_rknEns
ID__v_+0x26(804664c, fdd7c9e4, fe2f29f0)
libxul.so`__1cUmozJSComponentLoaderKReallyInit6M_I_+0xdc(8154338)
libxul.so`__1cUmozJSComponentLoaderKLoadModule6MpnMnsILocalFile_ppnJnsIModule__I
_+0x11b(8154338, 8154228, 80468c8)
libxul.so`__1cWnsComponentManagerImplVAutoRegisterComponent6MpnMnsILocalFile_rnI
nsTArray4nODeferredModule___i_I_+0x2b4(8111678, 8154228, 80469c0, 0)
libxul.so`__1cWnsComponentManagerImplWLoadLeftoverComponents6MrnKnsCOMArray4nMns
ILocalFile___rnInsTArray4nODeferredModule___i_v_+0x89(8111678, 80469bc, 80469c0
, 0)
libxul.so`__1cWnsComponentManagerImplMAutoRegister6MpnHnsIFile__I_+0x354(8111678
, 0)
libxul.so`NS_InitXPCOM3_P+0x55a(8046e44, 8090438, 8046ba4, fe2ee038, 33)
libxul.so`__1cSScopedXPCOMStartupKInitialize6M_I_+0x4b(8046e44)
libxul.so`XRE_main+0x14ae(1, 8047ce4, 80903f8)
main+0x769(1, 8047ce4, 8047cec)
_start+0x7a(1, 8047da4, 0, 8047daf, 8047dbb, 8047dcb)
> 
[stevel@grommit:dist] 557$ 

sb on os @ shdh, part trois

Sunday, Jan 6. 2008  –  Category: OpenSolaris, Songbird

[stevel@grommit:dist] 509$ ./songbird
ld.so.1: songbird: fatal: relocation error: file /export/home/stevel/sb/trunk/compiled/dist/components/sbMetadataHandlerTaglib.so: symbol __1cDstdIios_baseJ_Loc_init2t6M_v_: referenced symbol not found
Killed

okay - so it didn’t quite segfault.

still.

disappointing.

sb on os @ shdh, part deux

Saturday, Jan 5. 2008  –  Category: OpenSolaris, Songbird

whew. wow. that was messy.

songbird builds successfully now w00t.

i’m debating about whether i should just call it a night and head home feeling good, or if i should go ahead and try running it only to be disappointed by the inevitably segfault.

sb on os @ shdh

Saturday, Jan 5. 2008  –  Category: OpenSolaris, Songbird

(that’s songbird on opensolaris at super happy dev house)

got here about an hour and a half ago. so far my numbers are: 1 coffee consumed 1 coke consumed 2 beers consumed 1 songbird repo checked out 1 xulrunner repo checked out

and more importantly 1 xulrunner binary built (thanks to triskelios’ patch - and some patch cleaning up since some of the SB patches have been committed to recent xulrunner trunk)

i never thought seeing:

[stevel@grommit:bin] 505$ uname -a
SunOS grommit 5.11 snv_75 i86pc i386 i86pc
[stevel@grommit:bin] 506$ ./xulrunner
Mozilla XULRunner 1.9b3pre

Usage: xulrunner [OPTIONS] xulrunner APP-FILE [APP-OPTIONS...]

would get me so excited.

SHDH: Getting Songbird to sing on OpenSolaris

Thursday, Dec 27. 2007  –  Category: OpenSolaris, Songbird

Contrary to much popular opinion, I’m not actually employed at Songbird to work on porting it to OpenSolaris… so I haven’t had any time to really work on it at all. Fortunately, next weekend looks free, and coincides nicely with SuperHappyDevHouse, yay!

So with that, I’m planning to be at SHDH 22 next weekend Jan. 5th… if anyone in the SF Bay Area wants to meetup to hack on getting Songbird running on OpenSolaris, t’would be grand. If you’re not in the area, we should try and coordinate via IRC. I know there are at least 3 people (Triskelios, Ken, & dynamicproxy) who have given this a shot already - so let’s try and coordinate efforts and get this thing singing.

php5 “Unable to fork” on Solaris

Thursday, Dec 6. 2007  –  Category: Grommit, OpenSolaris

I was battling with inconsistent behaviour for grommit with PHP5 periodically (50% of the time, roughly) throwing the dreaded “Unable to fork” error (seen on both exec() & passthru() calls). Googling for it gives tons of great info for what causes this problem on Windows; but virtually none on Solaris.

Finally after some truss and DTrace-fu, I saw what was happening. exec() was calling open(), and since httpd is a 32 bit process, it was hitting the 256 file descriptor limit.

Booooooo.

A quick modification to /lib/svc/method/http-apache2 to do a:

ulimit -n 65535 LD_PRELOAD_32=/usr/lib/extendedFILE.so.1 export LD_PRELOAD_32

before the line calling apachectl at the end fixed the problem for me.

Seems like other people should have hit this problem though; which makes me curious why I don’t see any other hits on Google about it.

a dearth of packages

Friday, Nov 23. 2007  –  Category: Grommit, OpenSolaris

This Thanksgiving I give thanks for all the packaging work underway in OpenSolaris. I look forward to a day in the future where I don’t have to spend so much time building out packages for all the updated software I need.

I’m building out the new Grommit machine, and part of what Grommit hosts is a whole bunch of Zones. I like to try to keep everything clean with pkg installs, so I’ve spent a good part of the past two days building Solaris (snv_75a) packages for Dovecot, PHP5, and Postfix.

Yes, I realise that packages for these already exist on Blastwave and elsewhere, but I want ones that leverage the Apache2, MySQL, and OpenSSL that Solaris already ships with. Anyway, thankfully the software all built fairly straight-forward which made my life a lot easier.

About the only thing left that I need is SpamAssassin and I’m hoping to use Eric B’s package that he put together for the OpenSolaris mail server for that. Eric passed me the Postfix package they use but unfortunately I ended up needing to rebuild it anyway since I wanted it to have TLS + SASL and authenticate against the Dovecot SASL.

I’ve done migrations of all the ZFS data for the home and web directories (yay for zfs send | ssh zfs receive), and I’ve installed Mailman (which I didn’t bother putting in a package since it cleanly separates itself in its own directory). I did a test migration of the MySQL database (which seemed to barf only on a field named ‘key’, which fortunately was easily worked around by escaping it with backticks). I’m pretty happy with the general build out of the machine so far. When I built out grommit last time it was early on in my days of using Solaris, so in general I’d made a mess of installing software all over the place, some in Blastwave, some from source, and things poorly and sloppily done everywhere. I’m trying to do things properly this time to make it easier to quickly commission zones that don’t require a whole bunch of finangling to get running cleanly.

I wonder if I can get grommit migrated by Sunday. That’d be a nice Thanksgiving surprise.

nightingale is hatched

Tuesday, Nov 13. 2007  –  Category: OpenSolaris, Songbird

Project Nightingale has officially been hatched, and I’m psyched to be able to post this with both the Songbird AND OpenSolaris blog tags. Albert (aka Triskelios) has done some impressive work already in trying to get XULRunner and Songbird ported over to OpenSolaris. We’ve hit some speedbumps though, and thought that formalising it into a project and getting some more eyes and hands on things would help speed it along. We’ve been approved and are being graciously sponsored by the Desktop Community.

I’m really psyched and looking forward to having the most kickass media player (no bias whatsoever) ported to and running on the best operating system (again, no bias whatsoever). Being able to combine my two favourite open source projects rocks.

So if you’re a desktop hacker, a media player expert, a Mozilla whiz, or just someone looking for a fun project to hack on - please join the project, and help out. We’ll have Albert’s patches posted soon, and we’ll be having our discussion in the Desktop forum, or you can subscribe to the Desktop mailing list.

Closed processes and development in an open source world

Sunday, Nov 4. 2007  –  Category: OpenSolaris, OpenSource, Songbird

I just read an interesting article on Linux.com criticising two examples of poor showing in the open source world, one revolving around a KDE icon theme, and the other around GIMP’s UI team.

Certainly I see some parallels in any open source project, both OpenSolaris & Songbird included.

First: my opinions. You didn’t ask for them, but you’ll get them anyway. Blogs rule like that. For the two examples cited in the article, I actually have differing opinions.

For the icon example, I disagree with the Oxygen project and agree with the writer. The icon developers gave up any rights they had to restrict redistribution with their license they published under. The derivative theme was well within its rights to redistribute the Oxygen icons.

For the GIMP UI example, I actually agree with the GIMP UI team. I think they way in which they phrased it could have been better. Perhaps instead of “I am afraid that I do not have positions open at the moment.”, he could have expanded and invited the volunteer to submit his work for review and inclusion without necessarily being a part of the formal GIMP UI team. While some people may disagree with me, I think UI work benefits from having a core team of people with a shared set of goals and design aesthetic. Adding another team member could alter that dynamic and/or add more overhead to the team.

Anyway - these are pertinent issues to both communities in which I participate in. OpenSolaris perhaps more than Songbird - since Songbird has no internal/closed repositories of source. But one prime example of a project people complain about is installation, or packaging - and I know I’ll probably get heat for this, but I think doing closed development is not that bad. The packaging project took some heat for doing its initial prototyping and development (I know I had at least one debate with sch about it ;-)) between its team members before opening up and publishing its work. But given the conflict that can arise between people deriving work prematurely (e.g. the Oxygen case), or even just the issue of it being a design-in-progress (let alone a work-in-progress) means its often easier to be more agile and develop the core of the project between team members (whether that’s internal to Sun or involving external members).

Open source your work when you are ready - not when people ask you to

When you are ready for people to hack on your code, make derived works, and submit features and bugs - then you are ready for open source development. If you aren’t ready for those, then don’t. Open sourcing code isn’t free (assuming you care about your open source community - throwing code over the wall, as always, is cheap and easy - and you get what you pay for). If you want a thriving community then you have to be ready to spend time to cultivate it. If you’d rather be hacking on code to get something initially out - then you shouldn’t publish it initially.