/srv/irclogs.ubuntu.com/2011/03/30/#ubuntu-installer.txt

=== skaet is now known as skaet_afk
=== nevrax is now known as Guest62798
=== skaet_afk is now known as skaet
=== skaet is now known as skaet_
juzzy__hi06:06
=== skaet_ is now known as skaet_afk
mptev, I think these bugs are fixed in Natty:09:27
mptbug 16432609:27
ubot2Launchpad bug 164326 in ubiquity "ability to use existing partitions to install Ubuntu" [Wishlist,Confirmed] https://launchpad.net/bugs/16432609:27
mptbug 13108409:28
ubot2Launchpad bug 131084 in ubiquity "Partitioning language needs resimplification" [Undecided,Confirmed] https://launchpad.net/bugs/13108409:28
mptbug 15518509:28
ubot2Launchpad bug 155185 in ubiquity "More warning about installation" [Undecided,Confirmed] https://launchpad.net/bugs/15518509:28
evmpt: thanks! sorting now09:29
mptbug 29167709:35
ubot2Launchpad bug 291677 in ubiquity "partitioner in installer forgets settings when you go back" [Undecided,Confirmed] https://launchpad.net/bugs/29167709:35
mptor maybe not09:35
mpt(depending on whether it remembers the settings when going back after an error)09:36
mptbug 374913, fixed by the new design09:38
ubot2Launchpad bug 374913 in ubiquity "the graphical partitioner shows complete nonsense" [Undecided,New] https://launchpad.net/bugs/37491309:38
mptbug 44505309:41
ubot2Launchpad bug 445053 in ubiquity "installation is not user friendly" [Undecided,New] https://launchpad.net/bugs/44505309:41
mptI think bug 349479 is either fixed, or a duplicate of bug 28762009:43
ubot2Launchpad bug 349479 in ubiquity "[Jaunty] partitioner doesn't fit on screen with large number of partitions" [Undecided,New] https://launchpad.net/bugs/34947909:43
ubot2Launchpad bug 287620 in ubiquity "Partition diagrams are far too wide for the default installer window" [Undecided,New] https://launchpad.net/bugs/28762009:43
mptI assume bug 549432 is fixed, though I haven't seen it fixed myself yet. :-)09:46
ubot2Launchpad bug 549432 in ubiquity "Partitioning: unclear option with no reference" [Undecided,Confirmed] https://launchpad.net/bugs/54943209:46
mptbug 362588 seems fixed09:58
ubot2Launchpad bug 362588 in ubiquity "Installer presents confusing/misleading partition information" [Medium,Confirmed] https://launchpad.net/bugs/36258809:58
evreplied10:48
cjwatsonev: any thoughts on https://bugs.launchpad.net/bugs/735072?  seems like a fairly fundamental problem with the new hostname selection algorithm12:31
ubot2Launchpad bug 735072 in ubiquity "The hostname proposed by installer is too long for file sharing to work correctly." [Undecided,New]12:31
cjwatsonsee also https://answers.launchpad.net/ubuntu/+source/ubiquity/+question/15095912:31
* cjwatson notices that subprocess.Popen.communicate doesn't retry on EINTR (bug 745014), contemplates Python bindings for libpipeline, and then realises libpipeline doesn't retry on EINTR either ...13:01
ubot2Launchpad bug 745014 in ubiquity "keyboard-configuration.postinst died with SIGPIPE" [Medium,Confirmed] https://launchpad.net/bugs/74501413:01
cjwatsonstill, at least I can fix the latter13:01
evcjwatson: sorry, I was off helping the design team with natty screenshots13:03
evreading now13:03
evare we certain that this is a limit in the specification?13:05
* ev digs13:05
* ev sighs13:10
ev"This issue occurs because the NetBIOS domain name has a 15-character limitation."13:10
evhttp://support.microsoft.com/kb/226144 http://support.microsoft.com/kb/90926413:10
evcjwatson: truncate?13:10
evcjwatson: mpt suggested showing a different name for netbios only13:11
evso we could generate a stub smb.conf, I suppose13:11
evdoes that sound insane?13:11
cjwatsonI don't know enough about samba13:18
cjwatsonslangasek would be a good person to ask13:19
cjwatsonI wonder if we need to care about network uniqueness in the first n chars13:19
evI don't follow on the network uniqueness point, but I've asked slangasek in #ubuntu-devel about truncation13:24
cjwatsondon't we try to make sure that the hostname isn't seen on the network at installation time?13:24
evyes13:25
evseparate from this, but indeed13:25
cjwatsonwell, truncation will change the uniqueness properties13:28
evah, good point13:28
evI think we'll be okay13:28
* cjwatson contemplates just calling signal.siginterrupt after every call to signal.signal14:00
cjwatsonwhy oh why doesn't Python just let you set SA_RESTART directly (when available)?14:01
cjwatsonsilly language14:01
cjwatsonbut wait, we don't even install any signal handlers ourselves14:02
cjwatsonso I don't see where EINTR is coming from, unless Qt installs some signal handlers or something14:05
cjwatsonsomething is setting SIGCHLD without SA_RESTART in the ubiquity frontend process, so that's a possibility14:15
cjwatsonI blame Qt14:18
evgenerally a good approach14:20
cjwatsonI think it may actually be accurate here :-)14:21
cjwatsonQProcessManager does the thing above14:21
davmor2cjwatson: could it be you got too used to GTK commands and was spoilt :D14:22
cjwatsonvery tempting to have the KDE frontend do signal.siginterrupt(SIGCHLD, False)14:23
cjwatsonQt *used* to do SA_RESTART, but qt.gitorious.org doesn't go back far enough to show when that was deleted and why14:29
CIA-31ubiquity: cjwatson * r4626 trunk/ (debian/changelog ubiquity/frontend/kde_ui.py):14:44
CIA-31ubiquity: * KDE frontend:14:44
CIA-31ubiquity:  - Force the SIGCHLD handler installed by QProcessManager to use14:44
CIA-31ubiquity:  SA_RESTART, in the hope of avoiding problems with Python's lack of14:44
CIA-31ubiquity:  EINTR-safety (LP: #745014).14:44
CIA-31ubiquity: cjwatson * r4627 trunk/ (2 files in 2 dirs):14:49
CIA-31ubiquity: Show ext2 and btrfs in the same colour as ext3 and ext4, rather than14:49
CIA-31ubiquity: in the same colour as free space.14:49
evcjwatson: incidentally, I thought python handled eintr in subprocess now? http://hg.python.org/cpython/rev/6e664bcc958d/14:57
cjwatsonnot sufficiently15:02
cjwatsonthey only fixed part of the problem15:02
evperhaps it's worth replying to http://bugs.python.org/issue1068268 then?15:02
cjwatsonsomebody already sent a patch that fixed the bits that matter and they ignored it15:03
cjwatsonand there's another open python bug for fixing things further down the stack15:03
evI don't suppose you have a link to this?15:03
cjwatsonyes, I put my research in bug 74501415:04
ubot2Launchpad bug 745014 in ubiquity "keyboard-configuration.postinst died with SIGPIPE" [Medium,Fix committed] https://launchpad.net/bugs/74501415:04
evthanks15:04
cjwatsonusing SA_RESTART for all signal handlers is arguably safer anyway15:05
cjwatsonasking around locally, anything that uses stdio plus signal handlers without SA_RESTART is fundamentally unsafe15:05
cjwatsonyou can't retry stdio calls on EINTR since data may be buffered and you have no way to know this15:05
cjwatsonso unless the library does it for you (which AFAICS glibc doesn't), you're out of luck15:06
cjwatsonit's hard to see from python code at first glance whether it's using syscalls or stdio, but I think some of it may be stdio15:07
evcjwatson: to break away from the eintr conversation briefly, mpt and I are working through the case where ntfs is in an error state15:31
evwe're still working through the ui, but..15:31
evwhat are your thoughts on if ntfsresize --info fails, run ntfsfix, *ui goes here*, reboot15:35
evcjwatson: obviously that doesn't help massively when the error is actually at the block level15:36
evbut chkdsk might have slightly better ui for that scenario anyway15:36
evntfsfix> apparently sets the dirty flag, which I hope is enough to trigger chkdsk before mount in windows, but would have to test15:37
cjwatsondo all ntfsresize --info failures correspond to "run ntfsfix"?15:37
davmor2ev: it does15:37
davmor2ev: ntfsfix does some common repairs and then triggers chkdsk iirc from the man description15:38
evcjwatson: well, that's why I mentioned the block level error case.  Are there others you can think of?15:39
cjwatsonI'm not sure I know it well enough to make sure15:41
cjwatsonI think it would require going through the source15:41
evyeah, digging through it now15:42
evSA_RESTART> odd that ronaldoussoren mentioned it as a solution in http://bugs.python.org/issue9867 yet it wasn't taken (at least not from what I can see looking at the cpython3 source)15:44
=== skaet_afk is now known as skaet_
cjwatsonit's not something python can entirely enforce on its own15:47
evfair enough, I think my not-entirely-complete knowledge of the signals in question are preventing me from entirely understanding the correction solution15:49
cjwatsonyou need everything that installs a signal handler anywhere in the process to use SA_RESTART15:51
cjwatson(or you need to hack around it with siginterrupt)15:51
evright, but surely for the cases where that's python, it can handle it, no?15:53
cjwatsonpython's own signal handlers aren't problematic for this15:54
cjwatsonso that's moot15:54
cjwatsonfor the most part, anyway)15:54
evah, okay15:55
evthat's where I was getting tripped up15:55
cjwatsonit basically amounts to the SIGINT handler, which is likely to terminate the process anyway, and a few conditional things15:55
cjwatsonsetting signals to SIG_DFL or SIG_IGN doesn't count here, it's just explicit handler functions15:56
evugh, I need to spend more time with Stevens.15:58
evgod, the world before design specifications was a horrible place. Lets never go back.16:01
evhaving a single place to point at any time something is queried or contested, and most importantly can be changed there is wonderful16:01
evrather than having to trek back through bug reports, work items, wiki specs, ...16:02
cjwatsonI find a lot of it a bit invisible TBH16:02
evoh?16:02
cjwatsonI rarely know where to look to find that stuff16:02
evit should all be in the google doc16:03
cjwatsonwhat google doc? :-)16:03
evwhich I should actually link to on the wiki (and just just from a blueprint).  Fixes...16:03
evindeed :)16:03
evhttps://docs.google.com/a/canonical.com/Doc?docid=0AU5sFuLRpCpBZGZra2pqY2pfMTAxZ25rcnBnNXY&hl=en16:03
cjwatsonmemorable URLs FTW16:03
evI know, right?16:04
cjwatsonI guess when you aren't looking at it all the time ...16:04
artnayhi, is someone here in charge of ubiquity-debconf's strings/variables?16:05
cjwatsonthat doesn't really make sense as such, but just ask your question :)16:05
artnaytranslation template contains several strings with "${RELEASE}"16:05
artnaybefore maverick it used to be Ubuntu xx.xx but in maverick it was changed (just before freeze) to just "Ubuntu"16:06
artnayhow will it be with natty? just Ubuntu or Ubuntu <version> ?16:06
artnaythere are certain languages where one has to conjugate "Ubuntu" (for example, Try ${RELEASE} should be translated Kokeile ${RELEASE}a to be a proper sentence)16:08
artnaythere seems to be 9 strings which use this ${RELEASE} variable16:10
artnayhowever, if the variable won't be "Ubuntu" in natty, those translations will look plain wrong if conjugated16:12
artnayhopefully you understand why I'm asking this. it's all about small details, right? ;-) having amateurish translations with the 1st phase of installation won't give good impression16:14
cjwatson"Ubuntu", as far as I can see16:15
artnaycjwatson: got it, thanks.16:17
evcjwatson: google doc> I've added it to http://wiki.ubuntu.com/Ubiquity.  If you have any concerns with driving development of the installer from there, I would love to work to resolve them.16:24
evbug 745687 :(16:34
ubot2Launchpad bug 745687 in ubuntu "Brief text over the background image restarting after install" [Undecided,New] https://launchpad.net/bugs/74568716:34
cjwatsonyeah, I've seen that but it's a horrendous pain to debug live CD shutdown16:37
evindeed16:38
cjwatsonshall I stick it on plymouth for the time being, or do you think it's more likely to be ubiquity?16:39
cjwatsonI don't suppose we know ...16:39
mptartnay, depending on the circumstance, the partitioning step may say "This computer has Ubuntu 10.10 on it. What would you like to do?" -> "Upgrade Ubuntu to 11.04"16:39
evI'd stick plymouth on it, just for an expanded set of eyes16:40
artnaympt: good point, so maybe the version number should be included16:40
mptartnay, and "This computer has Ubuntu 11.04 on it." -> "Reinstall Ubuntu 11.04"16:40
evartnay: I'll stick comments in, but it's always Distro XX.XX where Distro is likely to be UBuntu and XX.XX may be 11.0416:40
evUbuntu*16:41
cjwatsonit's a bit odd that ${RELEASE} gets expanded to the distribution name16:41
artnaympt: that variable isn't used in those places16:41
cjwatsonat least as far as I could see from testing16:41
artnayhttps://translations.launchpad.net/ubuntu/natty/+source/ubiquity/+pots/ubiquity-debconf/fi/+translate?batch=10&show=all&search=%24{RELEASE} this is what I see (in Finnish)16:41
cjwatson>>> misc.get_release()16:41
evcjwatson: I don't follow. Is Ubuntu 11.04 not a release?16:41
cjwatsonReleaseInfo(name='Ubuntu', version='11.04')16:42
cjwatsonubiquity/plugins/ubi-language.py:306:        text = text.replace('${RELEASE}', release.name)16:42
cjwatsonetc.16:42
cjwatsonUbuntu 11.04 is; Ubuntu is not16:42
evoo16:42
artnayas I said, in previous versions it was with the version number, in maverick it was without16:43
evdear lord Unity, think you could run for more than 15 seconds without crashing?16:44
artnaympt: you're referring to this sentence: Install ${DISTRO} ${VER} alongside ${CURDISTRO}16:45
evjust checking through the code, one moment16:46
mptartnay, that string isn't specced to have ${VER} in it, though actually maybe it should16:47
mptartnay, https://docs.google.com/View?id=dfkkjjcj_101gnkrpg5v#4_5_1_Automatic_partitioning_o_847552608698606516:47
evcjwatson: # above Description lines will appear in rosetta, right?16:49
mptartnay, sorry, my mistake, that string is supposed to have ${VER} in it when there's an older version installed16:49
cjwatsonev: which ones?16:49
evI'd like to add a comment for translators.16:49
cjwatsonoh, I misparsed16:50
cjwatsonshould do, yes16:50
evokay, cool, thanks16:50
cjwatsonthough my rosetta memory is a bit rusty16:50
* ev goes off to clarify this in the code16:50
CIA-31pkgsel: cjwatson * r173 ubuntu/debian/ (changelog postinst):17:01
CIA-31pkgsel: localechooser/supported-locales only lists additional locales, so17:01
CIA-31pkgsel: calculate the set of language packs to install using the union of it and17:01
CIA-31pkgsel: debian-installer/locale (LP: #741304).17:01
artnayso, umh, to clarify. "Try ${RELEASE}" and "Install ${RELEASE}" are the problematic ones (in certain languages) that might need conjugation. should those have ${VER} as well? If yes, I could file a bug (to remind you devs)17:03
evartnay: hopefully that will help (it should end up on rosetta): http://bazaar.launchpad.net/~ubuntu-installer/ubiquity/trunk/revision/462817:07
CIA-31ubiquity: evand * r4628 trunk/debian/ (changelog ubiquity.templates):17:07
CIA-31ubiquity: Clarify substitution variables in the automatic partition templates17:07
CIA-31ubiquity: using comments.17:07
CIA-31pkgsel: cjwatson * r174 ubuntu/debian/ (changelog postinst): Fix Chinese special cases for language packs to work properly.17:08
evI'll add some more comments to cover RELEASE and friends now17:08
artnayev: that doesn't contain a fix/explanation17:08
artnayev: oh, ok. see 3.1 in mpt's link17:09
artnayUbuntu presents a list of languages with TWO OPTIONS ONLY: Try or Install Ubuntu.17:10
evsee my previous comment17:10
CIA-31ubiquity: cjwatson * r4629 trunk/ (debian/changelog ubiquity/install_misc.py):17:19
CIA-31ubiquity: localechooser/supported-locales only lists additional locales, so17:19
CIA-31ubiquity: calculate the set of language packs to install using the union of it and17:19
CIA-31ubiquity: debian-installer/locale (LP: #741304).17:19
CIA-31ubiquity: evand * r4630 trunk/debian/ (changelog ubiquity.templates): Clarify more debconf questions using comments.17:20
CarlFKThis should add to the installed /etc/apt/sources.list right?   d-i apt-setup/local0/repository string http://ddebs.ubuntu.com natty main restricted universe multiverse17:28
CarlFKI will swear it did, but even going back to my hardy installer it isn't.17:29
CIA-31pkgsel: cjwatson * r175 ubuntu/debian/changelog: releasing version 0.33ubuntu318:01
=== skaet_ is now known as skaet_otp
=== skaet_otp is now known as skaet
=== skaet is now known as skaet_afk
CIA-7ubiquity: cjwatson * r4631 trunk/ubiquity/frontend/kde_components/PartMan.py: typoes22:41
=== skaet_afk is now known as skaet
CIA-7ubiquity: cjwatson * r4632 trunk/ (bin/ubiquity debian/changelog):23:38
CIA-7ubiquity: Do a quick check at startup to see if the debconf database is locked by23:38
CIA-7ubiquity: something else, and exit slightly more gracefully than by raising an23:38
CIA-7ubiquity: exception (LP: #746020).23:38

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!