<PulkoMandy>
then we can carefully review the commits in-between and possibly find something suspicious
thomaslewis has left #haiku [#haiku]
<PulkoMandy>
following the "binary searching" instructions here: https://dev.haiku-os.org/wiki/BinarySearch you only need to test 4 or 5 of them to identify the problem
Maturi0n has quit [Remote host closed the connection]
<win8linux[m]>
Diver: For the latest revs, yes from USB. For 56010, from HDD on the desktop.
<win8linux[m]>
PulkoMandy: Yes that's what I meant, sorry. I cannot access it through the usual way of holding down the Shift key.
<Diver>
could be efi loader changes
MajorBiscuit has quit [Ping timeout: 480 seconds]
<win8linux[m]>
Currently downloading revisions 56018, 56030, 56082, 56107, and 56013.
<win8linux[m]>
Noticed that the download speed doesn't go beyond ~270-275 KiB/s for each file.
<win8linux[m]>
Is there a download speed limit enforced on the nightlies?
<win8linux[m]>
Both the desktop and laptop are AMD Zen+ systems, using 2600 and 2500U CPUs respectively.
<win8linux[m]>
Anyways while waiting on the revisions to finish downlooading, I could give some more details.
<win8linux[m]>
Grr Matrix, stop sending messages out of order.
jason123santaonirc has joined #haiku
<PulkoMandy>
we don't enforce speeds, but the hosting is on a single server that has limitations
jason123onirc has quit [Ping timeout: 480 seconds]
<PulkoMandy>
not sure if the usual places where opensource projects can host things for free would be happy with storing all the nightlies
<win8linux[m]>
Fosshost at least seems fine with that.
<win8linux[m]>
Shame that new applications are suspended at the moment.
<PulkoMandy>
that's what usually happen to anyone offering unlimited storage at a fixed price or for free
<Not-8690>
[haiku/haikuwebkit] pulkomandy 942dc09 - Use OpenSSL instead of GCrypt for web crypto support
x10z has joined #haiku
x10z has quit [Ping timeout: 480 seconds]
mmu_man has joined #haiku
tqh has quit [Quit: Leaving]
MajorBiscuit has joined #haiku
floof58 has quit [Quit: floof58]
floof58 has joined #haiku
Maturi0n has joined #haiku
<win8linux[m]>
PulkoMandy: Did some testing and 56013 (right after 56010) is the first rev to fail.
<win8linux[m]>
It only gets worse from there.
<PulkoMandy>
ok, did you re-test 56010 from USB as well?
x10z has joined #haiku
<win8linux[m]>
Not yet.
<PulkoMandy>
because the changes between 56010 and 56013 do not really look like they could affect booting in this way (we never know, but from a quick look at least), one possibility is that your current install works only because you have an even older EFI bootloader since that's not updated automatically
<PulkoMandy>
so we should make sure of that before investigating further
<PulkoMandy>
but if your existing install stopped working after an update, it's probably not that
<PulkoMandy>
is your video chipset an intel one?
<win8linux[m]>
Downloading that now and will test afterwards.
<win8linux[m]>
I'm using an AMD chipset.
<win8linux[m]>
Both the desktop and laptop are AMD systems.
<win8linux[m]>
There's more, however.\
<win8linux[m]>
By around 56107, the laptop starts getting stuck in a reboot loop.
<win8linux[m]>
Earlier revs would just end at a black screen.
<win8linux[m]>
At a black screen after the boot screen, specifically.
<PulkoMandy>
did you try failsafe video mode for these?
<win8linux[m]>
I cannot enter the boot menu on both systems by holding down the Shit key.
<win8linux[m]>
*Shift
<PulkoMandy>
shift doesn't work for UEFI, you have to spam the spacebar instead
<win8linux[m]>
Oh wow, what a typo. xD
<PulkoMandy>
I think we didn't find a way to detect the modifier keys state from UEFI
<win8linux[m]>
With the failsafe video driver on 56013.
<win8linux[m]>
* on 56013 on the laptop.
<PulkoMandy>
ok so that's the deskbar crashing, it seems
<win8linux[m]>
I'll try with some of the later revs.
<PulkoMandy>
ah, no, it's app_server in a thread communicating with deskbar
<PulkoMandy>
can you type 'bt' at that prompt?
<PulkoMandy>
we should get a bit more info then
<PulkoMandy>
I see you have FlatDecorator installed, that could cause problems too
<PulkoMandy>
if it wasn't updated to match the current internal APIs
<win8linux[m]>
That shouldn't be there, since the drive is imaged directly from an ISO.
<win8linux[m]>
It should be bone stock Haiku.
<win8linux[m]>
Didn't mention this before since I thought it was just due to using a very old flash drive, but there are times when a Haiku flash drive will show different revisions based on whether the system is using BIOS or UEFI.
<win8linux[m]>
This is a newer flash drive however, so it isn't caused by bad flash.
<win8linux[m]>
This can manifest even after wiping a flash drive and installing a newer Haiku rev on it.
<win8linux[m]>
The BIOS system would correctly reflect the new rev, but the UEFI system would still show the older rev that was supposed to be wiped.
<andreasdr[m]>
Good morning.
<PulkoMandy>
that means the efi bootloader isn't actually booting from the drive, but picking another haiku partition somewhere else
<PulkoMandy>
so we don't really know what versions you're testing then
<win8linux[m]>
On the laptop at least, there is no other Haiku partition.
<win8linux[m]>
Best possible explanation that I've got is that the EFI loader is loading from whatever data in the prior Haiku partition that hasn't been overwritten yet.
<win8linux[m]>
Meanwhile Bootman is correctly booting from the current Haiku partition.
<PulkoMandy>
well I don't know how you're writing to your USB media, it's possible if you don't fully erase it, and you have a partition near the end of the drive that isn't erased by the image occupying just the first few GB
<win8linux[m]>
I'm using dd.
<PulkoMandy>
yes so that will leave the end of the disk with whatever was there before
<PulkoMandy>
you can erase that to be sure: dd if=/dev/zero of=/dev/yourdisk bs=4M
<PulkoMandy>
that will write zeros all over the disk, then we're sure there's nothing left
<win8linux[m]>
Thanks, will do.
<bparker_>
you don't need to wipe the entire drive, just the partition table is plenty
<PulkoMandy>
no, because GPT has a backup of the partition table at the end of the drive
itanium686 has quit [Quit: Vision[]: i've been blurred!]
<PulkoMandy>
if that one is used for whatever reason, we can still find some of the old data
<PulkoMandy>
so let's be sure and wipe everything
<win8linux[m]>
Huh, did not know that GPT keeps a backup.
<win8linux[m]>
While the drive is being zeroed out, the EFI files in the desktop's Haiku install are dated to the 25th of March this year.
<Not-8690>
[haiku/haikuwebkit] pulkomandy 54c7910 - webcrypto: Implement HKDF in a way compatible with OpenSSL 1.1.1
<bparker_>
>just the partition table
<bparker_>
>no, the partition table
<bparker_>
um ok
<bparker_>
that's what I said
CatButts has quit [Ping timeout: 480 seconds]
<PulkoMandy>
well yes, but I don't know where it's located so erasing the device is simpler and good enough (in this case where it's an USB stick with just a few gigabytes at least)
<PulkoMandy>
if it was a few terabytes I would think more about it