IRC Logs for #crux-devel Tuesday, 2012-01-10

*** sepen has joined #crux-devel02:19
jueRomster: many thanks, I missed that because I had a newer version of atk installed :(02:30
Romsteri tried bumping atk to 1.32.0 same deal what atk version did you use?03:09
Romsterand in any case it'll break other stuff too.03:09
Romsteri tried CFLAGS+=' -DG_DISABLE_DEPRECATED' but that broke glib in another area of compile.03:10
jueRomster: 2.2.003:39
teK_-O2 works03:46
teK_does -O3 on x86 fail?03:46
sepenyes, I said that on my mail03:46
sepenmy mail was on x86_6403:46
sepenbut I could try it here, let me some minutes03:47
sepenyeah, -O3 failed too03:48
teK_so the best thing would be to sed that out :)03:49
sepennice, I'll patch iasl's Pkgfile to avoid -O303:49
teK_awesome, thanks03:49
sepenat the end, we'll improve again! :D03:49
sepenteK_: thanks to you too03:49
teK_no problem03:49
jueteK_: keep in mind that we don't guarantee a port to build with other than the default flags03:53
sepenthat would do the trick: export CFLAGS="${CFLAGS/-O?/-O2}"03:54
sepenjue: what do you suggest to me? keep the Pkgfile as it is?03:55
jueit's a rule from Per, and he was right because we cannot try all the different settings03:56
Romsteri asked if oyu had some other flags set don't you think -O3 is not standard and you should know that will break on some ports.03:56
teK_but if we find a broken one we should fix it...03:56
sepenwe? or code developers03:56
Romsterthat's like me saying i got about 12 ports that break when i enable LTO03:56
sepenmaybe I should mail 'iasl' devels03:56
juehmm, don't think so, where should we start here?03:56
teK_or that sepen03:57
Romsterwhich is true should i be adding that to the Pkgfiles to filter them too?03:57
jueRomster: no03:57
sepenI'm completely ok to add extra info (like not default confs) in README files03:57
Romstersame in teK_'s case to.03:57
Romsteri have thought of a directory for for adding files to adjust flags on each Pkgfile but i don't know if crux would be interested in such a change.03:58
jueRomster: I guess, you know the answer already ;)03:59
Romstersimplest fix is to either remove -Werror in the port or alternatively copy the port locally and force it to -O203:59
Romsteryeah i think it's not keeping to KISS.03:59
Romsterand besides -O3 is not always faster it can be slower too.04:00
Romsteronly set it on ports that can really benefit from it.04:00
juegenerally, there are a lot of other possible additions/changes to *FLAGS, if we start changing the ports for every possible break we will never finish04:00
sepen+1 jue04:01
Romsterif you want a speed up use -flto i already got a head start on it.04:01
sepenjue: but what should we do if we have enough info to avoid future problems to users?04:01
Romsteronly thing i think is ok too add is -fPIC where needed.04:01
Romsteryou could jsut unset CFLAGS in iasl? if it's a bit problem or filter everything but -march04:02
juesepen: which problems?04:02
Romsteranyways i've said my bit on this topic. :)04:03
sepenjue: for example, this case, we know that -O3 causes an error for this port, so we have some info that could be recycled to advise future users04:04
Romsterwiki page known to break ports?04:04
sepenpfff lot work04:04
RomsterteK_, can make an entry?04:04
Romsterthat's what wiki pages are for, are they not?04:04
sepenRomster: yep, but every port is a unique world04:04
juesepen: to be consitent you have to recompile the whole ports tree to see which port breaks with -O304:05
sepenRomster: at the end will have 1 page/per/port?04:05
sepensepen: yep, agree04:05
Romsterwho knows04:05
sepenjue: so README files?04:05
sepenRomster: and note that not all users have a browse-inet-connection when handling with ports/updates/etc.-04:06
sepenI discard wiki when working on production servers out of my desktop office04:06
Romstersorry i just assume everyone is on the net these days.04:06
Romstereven my phone has a browser and 3g04:07
sepenRomster: I'm on the net for those things, but I used 'portdbc' instead of portdb, for example04:07
juesepen: I vote for nothing, I guess we have more important things to do than adding README for ports that break with omg-flags04:07
sepenRomster: all I want to need is $ make -k <word>04:07
sepenjue: well04:07
juebut it's up to you, oc04:07
sepenjue: maybe we should update pkgmk.conf(5) with more info about 'export CFLAGS=STRING' sections04:07
sepenjust we need to say that '-O2' is safe to work04:08
sepenor is it noted in another place?04:08
Romsteradding README file to ever port would mask the important README files.04:08
juewell, it's more a handbook thing04:08
sepenbut pkgmk.conf is the place where change flags04:08
Romsteri take it for granted every time i edit the flags in pkgmk something will break04:09
Romsterall experienced users should know that?04:09
juesepen: we are a DIY distro and a user should know what he is doing if he change the defaults in pkgmk.conf04:10
Romsterhaving said that it may not hurt to add to the handbook that custom flags wont work on all ports04:10
Romsteri actually agree with jue for once on that one.04:10
sepenjue: yep04:10
jueadditions/corrections to the handbook are always welcome04:11
Romstermaybe pkgmk can be edited a little to show cfl;ags ldflags cxflags cppflags anything like that before it runs build() so when we ask for the log we will know why.04:12
Romsteras not all ports are so verbose.04:12
Romsterat compile04:12
sepenRomster: # set -x, since pkgmk.conf its included as source file it should work04:13
sepenor define export CFLAGS="......." && echo "CFLAGS: $CFLAGS" in your pkgmk.conf04:14
sepenDIY one more time04:14
RomsterteK_, if oyur keen on editing each ports flags to omg-optimise you might be keen on my own modification, i had that feature in hvl pkgutils but it was pretty ugly with PKGMK_NAME_CFLAGS= etc in pkgmk.conf but i woukd't mind a /etc/pkg/name.flags file to fine tune as needed.04:14
Romsterbut i know it's not KISS so i wont expect crux to adopt that idea.04:15
sepenRomster: can't you edit pkgmk.conf for that?04:16
Romsternot really enough man power to focus on everything :/ should be focusing on when other ports catch up to glib and when glibc is stable and all the fixes that gcc.4.6.x is going to need on broken ports due to being more strict.04:17
Romstersepen, perhaps i haven't got around to doing it yet.04:18
Romsternot high up in my priority04:18
sepenRomster: from pkgmk.conf you can access to lot of env vars, like $name $version etc. the ones from Pkgfiles04:18
Romsteri'd prefer to profiles some ports that chew cpu and measure better performance.04:18
Romsteryes i'm aware.04:19
sepenso you could use something like '. /etc/pkg/foo.whatever' as you need it04:19
sepenand keep the system KISS04:19
sepenjust its your decission, for that I like CRUX04:20
sepenand for the same I dislike (for now) rc-ng04:20
juesepen: I think that's a different story ;)04:21
juewith rc-ng you have much more possibilites to tune the system as with our current one04:23
jue... and the whole raid/md thing is currently more or less broken04:24
juebut I agree that rc-ng is much more complicated04:25
juewhich in turn doesn't hurt me if we can ensure maintenance of it04:27
juesepen: do you have more detailed objections?04:29
teK_to me it's utter crap if I cannot use -O3 in pkgmk because of a single port04:36
teK_the best solution would be to make upstream fix this04:36
jueteK_: what's your opiniom wrt rc-ng?04:42
teK_I have yet to test it04:43
teK_will do this afternoon :)04:44
jueyeah, please do it :)04:44
teK_then I'll let you know. wrt speedup I guess I won't see any since I'm using a ssd nowadays04:44
teK_Is tehre a wiki page with criticism / hints for rc-ng?04:45
teK_I have to admit that I did not follow the ML closely04:45
jueno, only the thread on crux-devel04:45
teK_so I'll read that first04:45
teK_hunting food now04:48
jueTBH, I'm a bit amazed that nobody else tries rc-ng without beeing 'forced' by me04:50
Romsteri wasn't really forced i just got onto it when you reminded me. after i moved back to crux because of multilib.04:54
Romsterbut i can't speak of the others.04:54
sepensorry I had a meeting with my bosses05:08
pitillojue: the best feedback will be the one provided over raid/md setup. In a user enviroment with splitted partitions is working right. And about difficult, I think it's more complex but it provides better alternatives for other scenarios than the current one05:32
Romsteri'm using lvm on top of raid no isses on rc-ng but i don't think i had any issues starting on the normal rc either.05:33
pitilloabout CFLAGS, IMHO if they aren't fixed upstream, why make it harder than overriding/fixing (forcing the use of the default ones) them in the current broken port until it's fixed upstram (the maintainer will notice if it's or isn't fixed upstream)05:33
juepitillo: where is the border here? should we fix everything that comes like omg-opti-flags or only -O3?05:35
pitilloI can't provide any feedback about that setup. My scenario is using 4 disks in vbox with partitions splitted there (/,/home,/usr,/var) if I remember right05:35
pitillojue: I think the maintainer should provide a working port. This means it should work with standard CFLAGS. The omg flags are for advanced users, which I think they will rebuild the port without those flags05:36
juepitillo: that's exactly what I've said above ;)05:37
pitillothe real problem begins if upstream uses those kind of flags, not users IMHO05:37
pitilloyes, IMHO your are right jue05:38
jueand as sepen said, it's really easy to extend pkgmk.conf with some logic to change FLAGS on the fly05:40
juee.g. the -x option of grep is very nice to do a reliable file-lookup, with that it's easy to keep a list with ports that need special flags05:43
jueor something like that ...05:43
acruxjaeger: only now with firefox-9... this is fast enough on ppc too and it can be used but there are some issues (no time to investigate)  building gnash and icedtea-2 thus i'm still with 3.6.x06:31
acruxthere are my working ports (xulrunner+firefox)06:33
acruxwith already packages for ppc3206:33
Romstercant' run firefox as 64bit?06:36
frinnstacrux: i still have your bugreport in my bookmarks :)06:36
Romsterff in 64bit seems ok here.06:37
frinnsthavent had time to really look into it (well, i have had the time. but theres always something else i'd rather do) :)06:37
frinnstsince thunderbird now supports a separate xulrunner library i'll revisit the issue again06:37
Romsterlike procrastinate06:37
acruxfrinnst: np :)06:37
acruxRomster: it works fine on 64bit too06:38
frinnstyou can probably just bump your firefox/xulrunner ports manually without any issues if you want newer versions06:38
Romsterdunno if there is flash for ppc06:38
Romsteryou probably sing gnash or something06:38
acruxthese specific packages in mozilla-9.tar.bz2 are built on ppc3206:39
frinnstRomster ^^06:39
frinnstholy fuck, the bug was filed almost a year ago06:39
frinnstthought it was sooner than that06:40
acruxRomster: i use gnash but i'd like to debug on ppc lightspark too06:40
acruxbut no time :/06:40
Romsterah hmm that's a issue.06:40
Romsteri wasn't aware of that one06:41
acruxfrinnst: please revies these ports: mozilla-9.tar.bz206:41
acruxdamn.. dislessic06:42
Romsterissue is will it even build with the included xulrunner the whole reason firefox is using a interneal one as the external xulrunner was holding firefox back from major changes to xulrunner that will break other programs that depend on xulrunner06:42
Romsterdyslexic and i'm glad i'm not the only one.06:43
acruxanyway it's everytime a pain build ff06:43
acruxRomster: :D06:43
Romsterwonder what gentoo does for gnash06:44
acruxbah.. gnash needs to be rebuilt everytime you change boost and/or ffmpeg06:46
Romsterhmm gentoo no gnash but i see xulrunner and gnash in frugalware it's doable to port it over to crux.06:46
acruxit's really a pity situation..06:46
Romsterboost doesn't get changed alot.06:47
Romsterffmpeg on the other hand somewhat more often.06:47
Romsteri want to go to the new API on ffmpeg but it breaks stuff.06:47
Romsterit would be best to revive xulrunner link to that than to rebuild ffmpeg and try to use it's internal xulrunner.06:48
acruxa good amount of tasks :)06:49
acruxi go for lunch, see you later06:49
Romsteryou got everything else besides xulrunner?06:49
Romsterk have a good lunch06:50
*** deus_ex has joined #crux-devel08:37
*** deus_ex is now known as pedja08:37
jaegeracrux: thanks for the info08:40
jaeger@seen rotwang08:45
clbjaeger: rotwang was last seen in #crux-devel 1 year, 11 weeks, 6 days, 20 hours, 54 minutes, and 30 seconds ago: <Rotwang> I see few ports in contrib that could benefit from it08:45
jaeger@seen #crux rotwang08:46
clbjaeger: rotwang was last seen in #crux 1 week, 0 days, 19 hours, 24 minutes, and 56 seconds ago: <Rotwang> tilman: it seems cmake is indeed shitty08:46
Romsterhmm that's a long time to not be in this channel08:59
sepen1 week is long time? I don't think so09:59
sepen@seen #crux pil10:00
clbsepen: I have not seen pil.10:00
sepen@seen vektori10:00
clbsepen: I have not seen vektori.10:00
jaegersepen: I believe he was referring to the 1 year 11 weeks10:08
sepenyes yes, anyways 1 year OR 1 week = 1 week ago10:19
*** sepen has quit IRC10:29
Romsteri was referring to the 1 year, 11 weeks, 6 days, 20 hours, 54 minutes, and 30 seconds ago18:38
Romsterwhich be more now.18:38
*** mavrick61 has quit IRC21:31
*** mavrick61 has joined #crux-devel21:32
jaegergrr... can't make grub2 find its damn config when booted via UEFI22:16

Generated by 2.11.0 by Marius Gedminas - find it at!