IRC channel logs

<damo22>Pellescours: i saw you sent in a patch to netbsd, but did you really need to send them the configure script? its 3000 lines of change

<Pellescours>I followed the comments in the configure.ac

<Pellescours>this is weird to me to, I tink they should remove the configure and let users do the autoreconf command thereself (or integrate the step in build.sh)

<Pellescours>I made progress wit ata compilation stuff

<Pellescours>I’m able to compile hurd with both ide and sata

<Pellescours>my problem is that now it panic at devw_attach

<damo22>hmm

<damo22>do you have a specific error message?

<Pellescours>the devw_attach("wd"...) failed

<Pellescours>I think it’s my ioconf generation which is incomplete for the wd part

<damo22>i think you can refer to a wd* in the ioconf even though its implemented in a different library

<damo22>since it will be present at runtime when linked

<Pellescours>Yes I can, this is what I did

<Pellescours>but it’s the ioconf.c that miss some config part for wd*

<damo22>libwd should have it

<damo22>what libs do you have now

<damo22>libpcidev_wd or something

<Pellescours>I compile hurd with librumpdev_ahcisata librumpdev_piixide librumpdev_ata

<Pellescours>the last one being the one I added to have the rumpns_atabus_cd and rumpns_wd_cd struct being in

<damo22>so librumpdev_ata has the wd elements and cd?

<Pellescours>yes it has

<damo22>you probably need to remove wd*.c and cd*.c from the ahcisata and piixide libraries and move them to ata

<damo22>if you havent already

<damo22>it should be makefile foo

<Pellescours>I already did

<damo22>ok

<Pellescours>it’s an ioconf problem

<damo22>maybe you can add them in as extern to the ioconf

<damo22>instead of having them missing

<damo22>im guessing here as i cant see what you did

<Pellescours>I will try to keep my changes and compile hurd with only one lib, just to see

<Pellescours>What I did, was to compile ata*.c files in libata + move the common part of *.ioconf in libata

<damo22>its a heirarchy as i understand: ahci/ide controller -> ata -> wd/cd

<Pellescours>right

<damo22>we need everything unrelated to the controller put in libata

<Pellescours>That’s what I did, and it worked, it’s just that at runtime, it’s not able to find th wd

<damo22>ok

<damo22>yeah it sounds like the ioconf

<damo22>but isnt the cd attached via a different lib? ahci/ide -> atapi -> cd

<damo22>so you need libs for wd and cd

<damo22>otherwise wd will clash twice when you compile in cd support?

<damo22>scsipi*

<Pellescours>wdc is in libata, so it won’t conflict, for scsipi, it’s just that libscspi will need a patch to use libata. But for now cd is not enabled in code, so I don’t care

<damo22>yes it is

<damo22>libscsipi is compiled in and cd works on rumpdisk currently

<Pellescours>But it’s not used If I try to boot with on wd0

<damo22>if you add a -cdrom iso in qemu it works

<Pellescours>yeah I see. I’ll check that after

<Pellescours>for now cd is not a problem to me, my problem is wd

<damo22>sure

<damo22>the autoconf was mismatched version, thats why you got a huge diff

<damo22>RE: netbsd patch

<Pellescours>yes, I know that my autoconf version missmatched, but I used the version installed on my linux (archlinux). Old version is 2.69, and mine is 2.71

<Zopolis4>arch should actually be easy for that

<Zopolis4>just use autoconf2.69

<Pellescours>I’ll re-submit the patch then

<Pellescours>damo22: I’m able to boot with ahci by compiling with libata (but without piixide)

<Pellescours>but with libata, you can’t build rump with only piixide (without ahcisata) because there is a missing symbol comming from sata_subr.c

<Pellescours>The panic happen when I compile with piixide and ahcisata but not ahcisata alone (I did not tried piixide alone due to missing symbol). I think panic is due to 2 modules trying to probe wd at the same time

<Pellescours>shouldn’t ata and sata being separated (sata = wd and ata = sd)?

<Pellescours>time to bed

<damo22>both wds cant be connected to the same controller so that should not be the problem

<damo22>unless its a naming problem

<damo22>ide should be wd, sata should be sd

<damo22>most likely

<Pellescours>so should I try to do the changes ?

<Pellescours>or is it probably something else ?

<damo22>perhaps sleep on it

<damo22>i have non-coding tasks to do now

<Zopolis4>oh wow a lot of packages with large reverse dependencies are just behind bad ifdefs

<Zopolis4>ok wait no theres only 6

<Zopolis4>and ive dealt with 3

<Zopolis4>even still

<youpi>Zopolis4: you can see that easily in graph-top.txt indeed

<Pellescours>damo22: by looking to the code, the call fails because the 2 modules asks for the same device, I will change the call to make them try attach fallback on lookup

<Pellescours>I DID IT

<Pellescours>the solution was simple, move the content of ahcisata_component to libata and make ahcisata_component and piixide_component being simple_component

<Pellescours>YAAAAYYYY, I’m able to boot the same disk in ide or sata

<Pellescours>now I have to create patches

<Pellescours>dumb question, how do I generate patch for quilt? I have commit of changes and I want to export the patches to put there content in debian/patches

IRC channel logs

2022-09-23.log