| Summary: | Tracker for all 2004.3-test bugs (all platforms?) | ||
|---|---|---|---|
| Product: | Gentoo Release Media | Reporter: | Chris Gianelloni (RETIRED) <wolf31o2> |
| Component: | Everything | Assignee: | Gentoo LiveCD Package Maintainers <livecd> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | dsd, genstef, greg_g, henrik, swtaylor |
| Priority: | High | ||
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | All | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: |
syslog for test4 with dopcmcia
check-udevstate-writable.patch mingetty-1.07.1.ebuild |
||
|
Description
Chris Gianelloni (RETIRED)
2004-10-17 08:36:45 UTC
install-x86-minimal-2004.3-test1.iso : irssi is missing on the install-x86-minimal-2004.3-test1.iso gpm doesn't work out of the box, as it did on former livecds. bootsplash working in silent mode, for verbose mode splashimage is set by the initscript lateron. switching to verbose leaves an artefact from the progressbar on screen. Roger install-x86-minimal-2004.3-test1.iso On my Pentium1 laptop I see: >> Mounting filesystems Illegal Instruction >> Attempting to mount CD:- /newroot/dev/cdroms/cdrom0 .... then it continues now to look for the offending thing I wanted to use taviso's http://dev.gentoo.org/~taviso/files/analyse-x86 but objdump is missing on the cd. Roger Roger: I'll look into the instruction problem as soon as I get a chance test-2:
- No Bootsplash on an IBM Thinkpad R40 (I didn't get them with the already running kernel, so I blame it on IBM/ATI)
- strange message from initrd (on said Thinkpad and my desktop box):
>> Filling filesystem...
>> Booting.mount: wrong fs type, bad option, bad superblock on /tmp/.initrd/dev,
or too many mounted file systems
*: Failed to move over the /dev tree!
..
INIT: version 2.84 Booting
- Splash Image (console background) pops up while the motd is displayed (thinkpad).
- display of motd takes two seconds until it fully appears (appears row-wise) (thinkpad)
I blame most of the errors as non-fixable because they seem to be related to the notebook hardware. The "wrong fs type" error however seems to be somehow related to genkernel.
That's it for now.
Oh no, wait : trying to boot in a Sun Ultra 5 failed - the prom said it cannot boot from that cd.
;)
Damnit, I knew I would forget something: irssi is missing. test3: I just hit enter on the boot commandline: irssi did a fine comeback! :) that's devfs starting not udev after keymaps I get: /dev/vc/0 no such device ? (paraphrased) "bootsplash" works fine on thinkpad 570 gpm still no worky out of the box but it tells me: * You need to setup MOUSEDEV and MOUSE in /etc/conf.d/gpm first I did that and after a manual start it worked just fine. dopcmcia NOT specified on commandline therefor I did: --------------------------- livecd root # /etc/init.d/pcmcia start Device 'i823650' does not have a release() function, it is broken and must be fixed. Badness in device_release at drivers/base/core.c:85 [<c0234bca>] [<c8c19dfd>] [<c0129de3>] [<c0105027>] * 'modprobe i82365' failed * Trying alternative PCIC driver: yenta_socket * Starting pcmcia... cardmgr[3141]: watching 2 sockets [ ok ]livecd root # --------------------------- pcmcia network card works fine then. Roger *** Bug 68358 has been marked as a duplicate of this bug. *** *** Bug 68246 has been marked as a duplicate of this bug. *** Created attachment 42307 [details]
syslog for test4 with dopcmcia
test4:
none already mounted or sysfs busy [oops]
otherwise: udev start looks good
the cool autostarting keymaps feature from test3 is gone, too bad.
gpm works now: yay!
starting pcmcia manually results in the above mentioned error messages but works.
gentoo dopcmcia spits out:
Autoconfiguring devices...
Done. [ ok ] * Not Loading APM
Bios support....
[ ok ] * PCMCIA enabled via cmdline...
* PCMCIA support detected.
* Starting pcmcia...
cardmgr[3107]: watching 2 sockets
cardmgr[3107]: could not adjust resource: IO ports 0xc00-0xcff: Device
or resource busy
cardmgr[3107]: could not adjust resource: IO ports 0x100-0x4ff: Device
or resource busy
cardmgr[3107]: could not adjust resource: memory 0xc0000-0xfffff: Inpu
t/output error
cardmgr[3107]: could not adjust resource: memory 0xa0000000-0xa0ffffff
: Input/output error
cardmgr[3107]: could not adjust resource: IO ports 0xa00-0xaff: Device
or resource busy [ ok ] * Starting input
hotplugging...
but pcmcia nic gets detected ok.
see attachment above
ungraceful shutdown: wants to write to /tmp/ udev something?
test4:
none already mounted or sysfs busy [oops]
otherwise: udev start looks good
the cool autostarting keymaps feature from test3 is gone, too bad.
gpm works now: yay!
starting pcmcia manually results in the above mentioned error messages but works.
gentoo dopcmcia spits out:
Autoconfiguring devices...
Done. [ ok ] * Not Loading APM
Bios support....
[ ok ] * PCMCIA enabled via cmdline...
* PCMCIA support detected.
* Starting pcmcia...
cardmgr[3107]: watching 2 sockets
cardmgr[3107]: could not adjust resource: IO ports 0xc00-0xcff: Device
or resource busy
cardmgr[3107]: could not adjust resource: IO ports 0x100-0x4ff: Device
or resource busy
cardmgr[3107]: could not adjust resource: memory 0xc0000-0xfffff: Inpu
t/output error
cardmgr[3107]: could not adjust resource: memory 0xa0000000-0xa0ffffff
: Input/output error
cardmgr[3107]: could not adjust resource: IO ports 0xa00-0xaff: Device
or resource busy [ ok ] * Starting input
hotplugging...
but pcmcia nic gets detected ok.
see attachment above
ungraceful shutdown: wants to write to /tmp/ udev something?
On my SMP/HT enabled laptop, the install-x86-minimal-2004.3-test4.iso throws up a root login prompt (which is obviously no good, since the password is scrambled). Pushing (or not pushing) F2 doesn't affect the outcome. I only tried test4 in vmware, but came over quite a few issues: These are not pretty: * Mounting sysfs at /sys... mount: none already mounted or /sys busy [ !! ] and usbfs already mounted or .. Maybe it would be nice to have RC_PARALLEL_STARTUP as autoconfiguring devices uses no ressources in vmware. The language selection does not work as the splash does not switch to verbose. Imo its the wrong way we are doing it, we should use lang=fr or lang=de and not ask any questions. And why is autologin not enabled on all consoles? Is there some local security issue if you can just switch vts on a livecd and become root? I'm sorry... but I didn't really ask for suggestions at the last friggin minute. Gentoo has done the keymap stuff this way for a while, and I don't intend on changing it right now. The sysfs and usbfs things are cosmetic and we're working to fix them now, but to be honest, they are not showstoppers and I'm willing to release with them there. We have root configured to autologin on every terminal, but it is not doing so (try a ps aux and you'll see mingetty --autologin on them all). Now, the real truth is that other than -releng, I get absolutely ZERO help in building the LiveCD releases. No matter how much I ask, I get almost no response from most of the Gentoo developers. Honestly, I was to the point of quitting today and still might simply because I am sick of working so hard and having very few people appreciate it. All I get instead is complaints and "honest opinions" on what *I* should change, but I see no attached patches, just bitching. I want to apologize for that last comment. I'm just really frustrated since we are way behind schedule and every time I fix something, another thing breaks. That doesn't include the fact that I am still putting in a 50 hour week at work. Hmm, i wrote here earlier but obviously forgot to hit post. I tested test4 on both my own machine and a friends. Bootup is fine, except for the /sys already mounted error. I had a quick look and I am not sure why this is, maybe kudzu is doing it? If I boot without pressing anything (i.e. not pressing F2 for verbose) it goes straight into the bash prompt. If I press F2 during bootup (of the gentoo kernel) then when all the initscripts have finished I am presented with a login screen. Similarly, the gentoo-nofb kernel always presents me with a login prompt. Will produce a quick baselayout patch to fix the saving udev state thing. Created attachment 42380 [details, diff] check-udevstate-writable.patch Against /etc/init.d/halt.sh See bug 68469 Created attachment 42384 [details]
mingetty-1.07.1.ebuild
Noticed that the latest mingetty in portage is 1.00 whereas I found an upstream
release of version 1.07. Does this help at all?
I have fixed the udev state thing locally by simply turning off that ugly tarball hack, since we don't need it anyway. As for the mingetty version, it very well could fix the problems we have been having. Another idea is to simply be dropped to bash --login rather than mingetty. I'm not sure if I see a point in using something like getty for the LiveCD and the functions only seem to be used on the LiveCD itself (they're in /sbin/livecd-functions.sh) Alright... I decided to use /bin/bashlogin (part of baselayout) instead and it appears to be working perfectly. Hi, I tried test4 on a A7N8X Deluxe r1 with a GeForce FX5900 GoldenSample by Gainward and there is no bootsplash, neither possibillity to boot with the framebuffer driver and resolution (ugly low rez black screen) I did that at a friend since i had this card and it didn't worked either with 2004.2, but I sold the card since, but if u need any info with it i'll ask her and report it. Both cd works with other cards (eg tried a TNT2, no problem) Actually... we don't use vesafb-tng simply because it is reported to not work on SMP boxes (our kernel is SMP-only), nor 64-bit boxes. Just tried booting the contents of install-x86-2004.3-test7.iso from my USB flash drive using cdroot=/dev/sda1 and failed. Previous livecds autoloaded the sd-mod module even when not using doscsi since it is needed for usb-storage, -test7 doesn't do this. Could autoloading of sd-mod be re-added? Btw, I've tried doscsi but it hangs my system when trying to detect tmscsim. -test8 adds the sd-mod mentioned in comment #24 - but still can not boot from USB flash device because of missing vfat support in kernel/initrd-modules. For that... you need to file a bug with plasmaroo to pull vfat into the initrd with genkernel... Done. Filed as bug #68619. -test6 and -test8 don't auto detect (and configure) my NIC anymore. The module used is "8139too". -test4 detected and correctly configured it using dhcp, with -test6 and -test8 it says nothing detected and I need to modprobe and dhcpd manually. Fortunatly, not a big problem, obviously. Re #22 and #23: if we don't use vesafb-tng, do you know if usual vesa has the same prob ? afterall it seems like windows/X11 can set refresh rates without any prob Normal vesafb *always* uses 60Hz, so it isn't a problem unless those GeForce cards simply don't like 60Hz. 68619 has NOTHING to do with this... please don't link non-relevant bugs to this one... it'll only serve to piss me off... especially since you waited until well after the QA period was over to start bringing this crap up. The test9 liveCD detected and configured my NIC properly again. There were also no warnings or errors during startup and the problem with the login prompt is also gone. Only dokeymap doesn't work. booting "gentoo dokeymap" doesn't ask for the keymap to use, the only possibility is to press "F2" during bootup, but then entering "10" (for de) doesn't seem to work. test9:
works nicely ;)
only issue left:
the silent splash doesn't show any progress at the beginning/ doesn't switch to verbose mode when asking for keymap input.
dopcmcia works fine as well.
(minor issue: while it looks for dhcp eth0 is assigned 192.168.0.2 which then vanishes as the dhcp request times out)
another minor issue:
'mount' output shows:
/dev/ROOT on / type xfs (rw,noatime)
I guess that comes from fstab somehow..
My thinkpad doesn't poweroff, but hangs at remounting local filesystems readonly. I believe this can be cured by the kernel config:
CONFIG_APM_REAL_MODE_POWER_OFF:
Use real mode APM BIOS calls to switch off the computer. This is
a work-around for a number of buggy BIOSes. Switch this option on if
your computer crashes instead of powering off properly. )
And I have tested on my pentium1 as well, no more illegal instruction anywhere :)
good job!
Roger
Well... 2004.3 is out the door so I think we can close this one... |