Lovechild | early question, are there any plans to implement reverse depencies in portage so that we can automatically remove orphaned packages? |
---|---|
mcp | Lovechild: mjc said, prelink is a very good thing. Can you confirm this? |
lostlogic | The problems with WOLK and some other kernels arise when the kernel source satisfies a package dependency that is normally satisfied externally (ALSA, dxr3, etc.) |
Lovechild | mcp: so far, it doesn't seem to work properly on my box, I get no speedup in load, and I loose all my icons |
mcp | Lovechild: oh |
mcp | lostlogic: maybe some `uname -r` dependencies :) |
Lovechild | mcp: but I've had it working once and it's jut amazing to see OpenOffice open up in less than a quarter of the time it took before |
lostlogic | mcp actually, I do have plans on how to deal with this, but am waiting on a portage feature that is in progress before I can impliment the plan ;-) |
mcp | Lovechild: wooo, incredible |
mattam | prelinking is just about speeding up startup for C++ applications, so AFAIK it is always good |
mcp | lostlogic: :-) k |
Lovechild | mattam: not if it breaks to much shit or causes instability |
Lovechild | but prelinking is a good thing(tm) |
mattam | right |
mcp | I just downloaded prelink, but first I have to same some of my libs first :) |
mattam | as always |
mcp | s/same/safe |
mcp | bah I hate these typos :) |
mattam | save maybe ? |
Method | prelink doesn't corrupt and is undoable |
Method | Nicholas: be sure and mention virtuals with java (and others) and java-config |
lostlogic | and as complained about before configuration file changes are updated using the etc-update tool... I think Nick is about to mention that.. |
lostlogic | Speaking of, does anyone familiar with it have comments on how etc-update could better handle file updates? |
seemant | be able to say "pff, just update everthing from x to y in the list" |
seemant | especially for things like postfix which has a gazillion updates, none of them in files you actually modify anyway |
seemant | I think it just updates its sample files |
seemant | all 500 of them :P |
lostlogic | seemant haha OK I gotcha, any chance you add that to my etc-update features bug so I don't forget about it? |
seemant | bug #? |
j2 | lostlogic: automatically update if the only part changed are headers |
seemant | maurizio: yes! that would be too cool too :) |
j2 | seemant: congrats for the talk :) |
lostlogic | j2 that is already done ... mostly ... |
seemant | thanks Maurizio |
j2 | lostlogic: oki |
lostlogic | j2 it catches if the only change is the cvs header line |
lostlogic | seemant 3804 |
seemant | thanks lolo |
j2 | lostlogic: maybe it should automatically delete ._cfg0000 if ._cfg0001 is available |
lostlogic | j2 I've pondered that before... ATM it checks all available updates for triviality and kills them off where they are trivial... but I spose it is usually preferable to kill all but the last file... |
lostlogic | j2 probably that becomse less of an issue if we do batch operation support |
j2 | indeed |
MrGrim | so who else is giving talks at this thing and on what topics? |
sarnold | http://umeet.uninet.edu/umeet2002/english/prog.eng.html |
MrGrim | thank you |
MrGrim | oh wow... that list is HUGE |
MrGrim | I wish I had known about it earlier... |
Arador | publicity on tv? ;) |
lostlogic | MrGrim everything is logged on that site somewhere |
lostlogic | wb seemant |
Nicholas | If anyone has any questions on things they'd like to see in portage, feel free to mention them. |
Lovechild | are there any plans to implement reverse depencies in portage so that we can automatically remove orphaned packages? |
MrGrim | Nicholas: I'd love to hear more details about any upcoming features... if there are any more |
lostlogic | Lovechild yes that is a coming feature hopefully he is about to mention it ;-) |
lostlogic | Lovechild already emerge --depclean -p will show you what packages are not in world nor depended upon by the world packages... |
Lovechild | Nicholas: How about support for things such as patching sources with diffs instead of getting the full tarball, and parallel download support? |
lostlogic | I'm still trying to get a patch that was submitted for portage to show the gentoo changelog entry for a package when updating it... |
Lovechild | lostlogic: now that would be a neat feature... |
lostlogic | Lovechild search bugzilla, someone already wrote a patch for it... that worked at the time... |
Lovechild | lostlogic: I know but as it wasn't picked up by portage mainline it was dropped I think.. same goes for parallel downloading |
lostlogic | Lovechild actually the parellel downloads and merges was something which Nick did impliment (messily) on a previous version of portage, and which should be coming back sometime after 1.4. |
lostlogic | Lovechild portage is in for complete abstraction and threading or so I hear... |
garoeda | Nicholas: how difficult is it to develop extensions to portage? |
lostlogic | garoeda what kind of extensions? |
garoeda | adding new features eg |
Lovechild | Are there any plans of moving portage to a database instead of the current system? |
garoeda | like implementing parallel ebuilds |
garoeda | Nicholas: or the reverse dependencies |
lostlogic | Lovechild the ebuilds will stay as they are, the dep tree and the installed package db may be moved to a DB but the flatfile method is definitely still prefered. (AFAIK, Nick, correct me?) |
mattam | what are Sticky Flags ? |
mattam | personal masks ? |
Method | sticky flags are flags that don't change when you upgrade the package |
Method | even if the global flags did |
Method | so that you get the same product back that you had before |
Method | but upgradeed |
j2 | useful for PHP for example .. |
Method | yes |
mattam | i'd like to have the packages i mask keep masked after an update. Is it that sort of thing or just not bumping revisions ? |
Method | you can use your own profiles |
Method | to hard mask stuff |
garoeda | Nicholas: thanks! |
mattam | right, but it's not updated at all if I make one |
lostlogic | We are currently setting up the capability for users to sync their trees via cvsup or sup so that local changes are merged as CVS does rather than overwritten. |
Lovechild | would it be possible to create a CFLAGS list that compiled some stuff with -O3 and crazy opts, and the rest with more conservative flags.. |
mattam | ok |
lostlogic | cvsup and sup apparently provide the speed of rsync with the merging capabilities of cvs (I've never used them personally tho) |
mattam | what about more interaction during the build process for devellopers/willing users ? |
Lovechild | Nicholas: nice short answer... :) |
lostlogic | mattam actually we work hard to keep all of our ebuilds non-interactive |
mattam | this may be a bad example but leaving alsa build all the drivers is not very good |
mattam | i said for _willing_ users |
lostlogic | seemant USE=interactive for mattam? :) |
mattam | :) |
seemant | I see |
lostlogic | mattam actually what we do for instance in alsa is it reads an environment variable 'ALSA_CARDS' tod ecide what to merge, if it is empty it does all |
seemant | mattam: what are the applications for that? |
Lovechild | people like myself have a special talent for breaking Gentoo, would it be possible to setup a FEATURE flag for debug, that block everything that's harmful for debugging - so we can provide good feedback to our beloved developers? |
mattam | seemant: perhaps for configuring things too |
Lovechild | lemme guess, sticky again :) |
Lovechild | Nicholas: did so already, no response on that one |
j2 | !! |
lostlogic | uh oh... *kicks nick's char* |
j2 | that's why I asked some time ago. a "debug" useflag |
lostlogic | don't we have one that most ebuilds ignroe? |
j2 | nope |
Lovechild | Nicholas: my bad, the bug it seems wasn't registered |
mattam | +1 for debug |
j2 | just several [ -z $DEBUG ]; then ... |
Lovechild | j2: would be nice if it was a feature flag, and then portage handled it, not the ebuilds |
Nicholas | Ok... Anyone else have questions or comments? |
lostlogic | j2 hmm... *takes a look* |
mattam | just not stripping is not enough |
mattam | portage already has a flag for stip/nostrip IIRC |
mattam | Nicholas: a feature request: add license to the small desc |
j2 | lolo: 9033 |
mattam | :) |
Lovechild | Nicholas: http://bugs.gentoo.org/show_bug.cgi?id=12272 bug filed as requested |
mattam | oh an having the ebuild replay only a part of the process, not restarting from the beginning each time |
mattam | right ? damn me |
Lovechild | I know rmerge2 does this, but resume support for emerge -e world ... there's nothing more annoying that hitting a bad ebuild combo after 2 says of recompiling |
mattam | clap clap clap |