Re: 3.0.18 testing needed
>>>>> "ADC" == Adam Di Carlo <email@example.com> writes:
ADC> I've uploaded i386 and powerpc boot-floppies to
ADC> The PowerPC ones are still uploading -- they vary in that they
ADC> enable i18n, which is not yet decided upon for powerpc.
ADC> The primary things to test for:
ADC> - no more flashing screen or nasty segfaults when dbootstrap
ADC> starts (If you do get the flashing screen on i386, please fall
ADC> back to trying another version,
ADC> - base installation actually works again (we got an unofficial
ADC> busybox/wget for these builds)
ADC> Testing here is crucial -- what I need to know is whether this
ADC> is a "plausible" release that generally works (despite the new
ADC> i18n) stuff. We'd like to release 3.0.18 within the next
ADC> couple of days.
ADC> Please send reports on success or failure to
ADC> firstname.lastname@example.org. Filing bugs against
ADC> 'boot-floppies', 'base-config', 'busybox', the kernel, etc are
ADC> also ok.
Ok, I tried out images-1.44/compact (from the 3.0.18-2001-12-14 dir) and
here is what I found:
- the language chooser dialog was too wide (as you already mentioned in
- a white screen flashed one time only just before the language chooser
dialog popped up.
- when selecting a swap partition and root partition, the chooser that
lists partitions on the HD never appeared. My test system was
initially partitioned as:
/dev/hda1 128M Linux Swap
/dev/hda2 3120M Linux type
When mounting a swap partition, /dev/hda1 was automatically chosen
and I was only asked if I really wanted to proceed with wiping out
hda1 and initializing it as swap. In other versions of b-f, I recall
a list being shown with only one partition (the swap type partition)
listed and automatically selected. This behaviour is new. Is this
intentional ? Or is it a bug ?
The same happened when I had to initialize a Linux
partition. /dev/hda2 was automatically chosen for me and I was only
asked about retaining kernel-2.0 compatibility and about the badblock
- occasionally, the screen would not clear and output from the previous
step would be interspersed with output from the currently executing
- When installing driver and kernel modules from a floppy disk
(driver-1.bin), I selected to install the net/tlan module as my test
system has a TI ThunderLan-based Compaq NIC. After I pressed enter to
insert the module, at the bottom of the screen I saw the following
message repeated 4 times:
/target/usr/sbin/modconf: cannot create : directory nonexistant
but the tlan module got loaded successfully, and I was able to
autoconfigure eth0 using DHCP. This was one step where after pressing
ENTER a few times to select and insert the tlan module the screen was
The base system was downloaded over the net - at one point it seemed
that busybox-wget got stuck tring to download debianutils (machine was
not hung, and network connectivity was OK). I hit Ctrl+C and restarted
the base system install and it seemed to continue fine after that.
Once base were extracted and installed, all of a sudden the screen
flashed white once and the language chooser reappeared. Checking one of
the debug consoles I saw the following message:
(none)user.info init: Process '/sbin/udbootstrap' (pid 1585)
exited. Scheduling it for restart.
After selecting English, the release notes appeared and then the "Make
System bootable" step appeared selected in the list of tasks. Selecting
it caused the system to loop over and over again and presenting me with
the language chooser, release notes screen, and "Make System bootable"
step over and over again. The same restart message appears repeatedly in
the debug console.
I will restart and retry the install all over again as this attempt has
ssahmed AT pathcom DOT com