Packages Repository

[jpmatrix]

@gen2: Are your packages compiled with optimization flags? In fact I find firefox and abiword in the feed take much longer time to start than under pdaX. Firefox takes 8 to 9 minutes to launch (!) and abiword takes a minute or two, while under pdaX it's 2 minute and a few seconds respectively. 
[div align=\"right\"][a href=\"index.php?act=findpost&pid=162404\"][{POST_SNAPBACK}][/a][/div]

IFWIW, firefox opens here in 3:45, abiword in 48 seconds...siag 7 (!) seconds...siag is pretty cool, btw, if you haven't checked it out, it's got an abiword-like thing in there in addition to spreadsheets and other stuff.
[div align=\"right\"][a href=\"index.php?act=findpost&pid=162415\"][{POST_SNAPBACK}][/a][/div]

i reinstalled openbsd from scratch and haven't yet reinstalled firefox .. however i remember it was taking more 3 than 8 minutes to load... i'll try asap.

anyway, i've compiled pidgin (and dependances) with optimizations listed and it seems to work great!
uploading all that to my repository right now!

next to compile : scrot ! thanks

[mathemajikian]

anyway, i've compiled pidgin (and dependances) with optimizations listed and it seems to work great!
uploading all that to my repository right now!

Going to have to try out pidgin. This could be a replacement for the pork application I'm currently using:  

Finch is the text-based version of Pidgin. It supports the same IM networks, but you can run it in a console window. You can use it on Linux, BSD, and other Unixes.

[jpmatrix]

anyway, i've compiled pidgin (and dependances) with optimizations listed and it seems to work great!
uploading all that to my repository right now!

Going to have to try out pidgin. This could be a replacement for the pork application I'm currently using:  

Finch is the text-based version of Pidgin. It supports the same IM networks, but you can run it in a console window. You can use it on Linux, BSD, and other Unixes.

[div align=\"right\"][a href=\"index.php?act=findpost&pid=162423\"][{POST_SNAPBACK}][/a][/div]

finch doesn't seem to be in the ports...

[mathemajikian]

finch doesn't seem to be in the ports...

It's built by default when you build net/pidgin. Look at pidgins pkg/PLIST and you'll see bin/finch.

[jpmatrix]

finch doesn't seem to be in the ports...

It's built by default when you build net/pidgin. Look at pidgins pkg/PLIST and you'll see bin/finch.
[div align=\"right\"][a href=\"index.php?act=findpost&pid=162435\"][{POST_SNAPBACK}][/a][/div]

whow ! you're right !
now i've to find the keyboard shorcuts

[gen2]

@gen2: Are your packages compiled with optimization flags? In fact I find firefox and abiword in the feed take much longer time to start than under pdaX. Firefox takes 8 to 9 minutes to launch (!) and abiword takes a minute or two, while under pdaX it's 2 minute and a few seconds respectively. 
[div align=\"right\"][a href=\"index.php?act=findpost&pid=162404\"][{POST_SNAPBACK}][/a][/div]

Just re-compiled abiword so we can have a psuedo-official comparison of flags:

Code: [Select]

export CPPFLAGS="-mcpu=xscale"
    export CFLAGS="-mcpu=xscale"
    export CXXFLAGS="-mcpu=xscale"
Abiword took 48 seconds to open.

Code: [Select]

export CPPFLAGS="-mcpu=xscale"
    export CFLAGS="-mcpu=xscale -mtune=xscale -O2 -pipe"
    export CXXFLAGS="-mcpu=xscale -mtune=xscale -O2 -pipe"
Abiword took 47 seconds to open.

Too close to even say there's a difference in this case, far as I can tell.

[ZDevil]

Thanks, gen2.
I check again with the output in terminal, it seems both abiword and firefox are complaining about pango errors. Will reinstall the packages and try again.
Still 47 secs is way too slow compared with pdaXrom. There abiword always launches in about 10 secs.
Just a note on the flag: Despite CFLAGS (for C compiler) and CXXFLAGS (for C++ compliler), CPPFLAGS seems never for optimization but simply for extra include paths.
Now I've got both my 16gb CF and Ambicom wifi CF working! Just started to play with the Port collections. Nice stuff!

[ZDevil]

@jpmatrix: I've finally got everything set up (CF wifi + 16GB internal CF) and had access  to the Port tree (really sweet  ). Along the way I've compiled and packaged new things based on gen2's newer library packages. What would be the best way to share them? If I upload them to your ftp, then they won't get along with the packages in your feed because of library issues. What do you think?  

[jpmatrix]

@jpmatrix: I've finally got everything set up (CF wifi + 16GB internal CF) and had access  to the Port tree (really sweet  ). Along the way I've compiled and packaged new things based on gen2's newer library packages. What would be the best way to share them? If I upload them to your ftp, then they won't get along with the packages in your feed because of library issues. What do you think? 
[div align=\"right\"][a href=\"index.php?act=findpost&pid=162539\"][{POST_SNAPBACK}][/a][/div]

just upload them, and i'll make a separate directory for you on matrixmen.free.fr
then one could choose to mix or not to mix
if you start tonight i'll let my computer awaken

[ZDevil]

Thanks. Now I think of an issue: even if I upload the new packages to your server, how about the dependency tracking? Because my packages are using gen2's libraries, pkg_add pointing to the new location won't be able to pull the dependency packages at the same time, or setting multiple ULRs in PKG_PATH will do (but how)?

[ZDevil]

Thanks, gen2.
I check again with the output in terminal, it seems both abiword and firefox are complaining about pango errors. Will reinstall the packages and try again.
Still 47 secs is way too slow compared with pdaXrom. There abiword always launches in about 10 secs.
Just a note on the flag: Despite CFLAGS (for C compiler) and CXXFLAGS (for C++ compliler), CPPFLAGS seems never for optimization but simply for extra include paths.
Now I've got both my 16gb CF and Ambicom wifi CF working! Just started to play with the Port collections. Nice stuff!
[div align=\"right\"][a href=\"index.php?act=findpost&pid=162448\"][{POST_SNAPBACK}][/a][/div]

An update:
After reinstalling the packages, both firefox and abiword load faster and run, without complaining about pango errors.
Abiword startup time between 30~40 secs. Typing fast or typing a long word will result in repetition of the same key strokes.
Firefox about 4 minutes. And response not very fast.
Anyone can confirm this?
Still double the time under pdaXrom. Any explanation?

[gen2]

Still double the time under pdaXrom. Any explanation?

Possibly the biggest factors there are Theo's development priorities, although I'm sure there's much more to the details than I'm able to illuminate. FWIW, those priorities are what keep this user interested...(well, that and: I have no doubt pdaXrom is ideal for many users, but my experience there led me to believe management was ad-hoc at best, developer shortages chronic, long-term outlook shaky, communication to user-base very not good. Just my personal opinion, not trying to scare anyone away from a speedy rom that works for them!)

I'm thinking the OpenBSD team is handling the distro, Uwe and others handle the Z port (much smaller relative effort), and that's what we're using...contrasted with other roms which are trying to maintain an entire distro/port...I don't know if that approach can scale over time vs. developer-base, especially as the Z is overtaken by new hardwares and interest wanes.

OpenBSD's performance and usability is occasionally criticised. Felix von Leitner's performance and scalability tests,[44] indicated that OpenBSD lagged behind other operating systems. In response, OpenBSD users and developers criticised von Leitner's objectivity and methodology, and asserted that although performance is given consideration, security and correct design are prioritised, with developer Nick Holland commenting: "It all boils down to what you consider important."[45] OpenBSD is also a relatively small project, particularly when compared with FreeBSD and Linux, and developer time is sometimes seen as better spent on security enhancements than performance optimisations. Critics of usability often point out the lack of user-friendly configuration tools, the bare default installation,[46] and "spartan" and "intimidating" installer.[47] These see much the same rebuttals as performance: a preference for simplicity, reliability and security; as one reviewer admits, "running an ultra-secure operating system can be a bit of work."[48]

http://en.wikipedia.org/wiki/OpenBSD

[ZDevil]

A short note on my porting:

I am now building a whole new bunch of ports under 4.2 via NFS@macbook, using a new user on Zaurus matching the gid and uid of the Mac user.
But then when I use this account to build stuff, the optimization flags are simply ignored, despite the fact that i have already put them in .profile, .kshrc (using ksh now) and even exported them explicitly.

But there is one solution, after reading the official FAQ section 15.3.3.  

#1 Simply create /etc/mk.conf (need root privilege of course)

#2 Drop these lines in the file:

Code: [Select]

CFLAGS+=-mcpu=xscale -mtune=xscale
CXXFLAGS+=-mcpu=xscale -mtune=xscale(It seems there is no need for -O2 and -pipe flags, as i notice generally the ports have these included in the makefiles.)

#3 Save and go ahead to port stuff. Now you get systemwide make optimization settings.

And a good thing is you don't need to reboot the system. The new settings take effect immediately.

[ZDevil]

A minor qualification:
Both the -O2 and the -pipe flags can be safely put together with the architecture optimization flags, because -pipe is simply to smooth the channeling during build and make can only see one (= the last) -Ox flag and ignore the others.

And a question (which is kinda obvious, but i am not absolutely sure):
Is Xenocara just the X stuff we find among the distribution sets?
If that's the case then there seems to be no need to (re)complie Xenocara, right?
Thanks.

And a call:
Is there someone being be so kind as to offer a small space to host my optimized packages for the current 4.2 beta? My 3200 now keeps running day and night rebuilding. The ready-made packages will save other users lots of time and troubles to go through the pain of builing themselves. Thanks in advance!

[mathemajikian]

And a question (which is kinda obvious, but i am not absolutely sure): Is Xenocara just the X stuff we find among the distribution sets?
If that's the case then there seems to be no need to (re)complie Xenocara, right?

If you've updated via the snapshots then you shouldn't have to.