IRC channel logs

2020-07-06.log

back to list of logs

<swivel>AlmuHS: if you seem unable to affect the logged output with your modifications, maybe check the md5sum of the Xorg binary and vesa_drv.so or whatever the driver is vs. the md5sum of the one you built to make sure it's reaching your test environment?
<youpi>what does your print line look like exactly?
<AlmuHS>The date of the file
<AlmuHS>pruebas@hurd:~$ ls -l /var/log/Xorg.0.log
<AlmuHS>-rw-r--r-- 1 root root 5494 jul 6 00:18 /var/log/Xorg.0.log
<swivel>you could also strings/grep the binary for something unique about what you're printing to ensure it's there
<AlmuHS>youpi: the first message is this xf86DrvMsg(1, X_WARNING, "VESA testing starts\n");
<AlmuHS>xf86DrvMsg(1, X_WARNING, "VESA testing starts\n");
<AlmuHS>others are like this
<AlmuHS>xf86DrvMsg(pScrn->scrnIndex, X_WARNING,"testing function %s\n", __func__);
<youpi>ok, that looks good
<youpi>swivel's idea to strings| grep the binary to make sure that your text is getting in is worth checking
<AlmuHS>hexdump?
<swivel>wouldn't be the first time someone pulled their hair out in such a scenario without even getting their changes in the execution environment for some unrelated reason ;)
<youpi>AlmuHS: no, strings should be enough
<youpi>yes, that's very common
<swivel>esp. if you're using package rebuild scripts and such
<youpi>non-beginners make the same mistake, they just remember to use such trick to check what they're doing wrong :)
<AlmuHS>then I have to compile manually, is not=
<AlmuHS>is not?
<youpi>no, that shouldn't be needed
<youpi>but first use strings and grep to make sure whether or not your messages are getting in what you have installed in the end
<AlmuHS>yes, but I have to find the binary first
<youpi>dpkg -L xserver-xorg-video-vesa
<youpi>will tell you the list of files
<AlmuHS>this? /usr/lib/xorg/modules/drivers/vesa_drv.so
<youpi>yes
<swivel>strings /usr/lib/xorg/modules/drivers/vesa_drv.so | grep "testing starts"
<AlmuHS>the string is not here
<youpi>how do you install your generated .deb file?
<AlmuHS> https://pastebin.com/bxRsyWA6
<AlmuHS>youpi: sudo dpkg -i xserver-xorg-video-vesa-dbgsym_2.4.0-2_hurd-i386.deb
<youpi>it's not the dbgsym taht you need
<youpi>it's the other one
<swivel>oops
<youpi>the -dbgsym one only contains the debugging symbols, not the actual binary
<AlmuHS>in other time I installed the real package. I've just notices this time might be not compiled well
<AlmuHS>wait a minutes to repackage
<AlmuHS>sudo dpkg -i xserver-xorg-video-vesa_2.4.0-2_hurd-i386.deb
<AlmuHS>pruebas@hurd:/usr/lib/xorg/modules/drivers$ strings vesa_drv.so | grep "testing starts"
<AlmuHS>VESA testing starts
<swivel>victory!
<AlmuHS>but, anyways, it's the only message that it's printed in the Xorg log
<AlmuHS>pruebas@hurd:/usr/lib/xorg/modules/drivers$ grep testing /var/log/Xorg.0.log
<AlmuHS>[555652.712] (WW) VESA testing starts
<youpi>at least you get something, that's a start
<youpi>which function is that?
<youpi>(really, use __func__)
<youpi>(and __LINE__)
<AlmuHS>I used __func__
<youpi>I mean everywhere :)
<AlmuHS>this is the initial function
<AlmuHS>VESAIdentify
<youpi>possibly VESAPciProbe is not getting called because of a libpciaccess call issue inside the Xorg server itself
<youpi>i.e. the xorg-server package
<AlmuHS>could be possible
<youpi>grep tells it's basically only in ./hw/xfree86/common/*
<youpi>that's worth checking
<AlmuHS>i can repeat the test recovering the 2017 libpciaccess, to see an ideal running
<youpi>that can be useful yes
<youpi>possibly you'll see functions getting called
<youpi>which would mean that the problem is not what these functions do, but that they aren't getting called at all
<youpi>thus the problem not triggering inside the vesa driver, but in the xorg server code calling the vesa driver
<AlmuHS>ok, wait a minutes
<AlmuHS>fuck, now it appears a segfault
<youpi>a segfault is a nice thing, it tells you where a problem actually got critical
<AlmuHS>the 2017 libpciaccess didn't have pci.
<AlmuHS>pci-arbitrer
<youpi>yes, sure, that's probably the trigger, but we need to know what exactly doesn't work with pci-arbiter
<AlmuHS>segmentation fault at address 0xc
<youpi>ok but theres also a backtrace, isn't there?
<AlmuHS>i think yes
<AlmuHS>wait, I'll copy Xorg log in pastebin
<AlmuHS> https://pastebin.com/TpW3vtXz
<youpi>perhaps you made a mistake in your message printing, that lead to the crash
<youpi>possibly pScrn is null
<AlmuHS>maybe I can replace It with a fixed value
<youpi>you can as well write 0 instead of pScrn->scrnIndex
<AlmuHS>ok
<swivel>pScrn ? pScrn->scrnIndex : -1 or something perhaps
<youpi>we don't really care about the screen number actually
<youpi>it'll just be 0
<swivel>cool, good practice to just be defensive in general WRT dereferencing pointers in debugging prints on unfamiliar code
<youpi>yeah
<swivel>almuhs: you've embarked on quite the toolbelt-enlarging rabbit hole, I'm just happy to see keithp's refactoring of Xorg into easily compilable modules is paying off in yet another situation :)
<swivel>it used to be brutal to iterate on something like this
<AlmuHS>the long functions and files is one of the older bad practices in coding
<damo22>youpi: you said that vm_allocate_contiguous can be in the kernel only, but we do need an RPC to access it from userspace
<youpi>yes
<youpi>I just meant that it doesn't need to be attached to the device master port, it can be on the host master port
<damo22>i see, i made it on the mem device
<youpi>because I don't think we'll ever need to redirect that in a bootstrap translator
<damo22>ok i will move it again
<damo22>to use host priv port
<damo22>im back to work now
<damo22>ttyl
<damo22>Hi AlmuHS does HS stand for HurdSMP ?
<AlmuHS>HS is a old alias
<AlmuHS>youpi: now Xorg works (using 2017 libpciaccess)
<AlmuHS>i go to check the log
<AlmuHS>youpi: now I can see my messages (and so much info) https://pastebin.com/a0Kw1iMQ
<AlmuHS>pruebas@hurd:~$ grep "testing" /var/log/Xorg.0.log
<AlmuHS>[557721.491] (WW) VESA testing starts
<AlmuHS>[557721.621] (WW) VESA(0): testing function VESAPreInit
<AlmuHS>[557721.621] (WW) VESA(0): testing in function VESAPreInit finished
<AlmuHS>[557721.911] (WW) VESA(0): testing function VESAMapVidMem
<AlmuHS>pruebas@hurd:~$ grep "libpciaccess" /var/log/Xorg.0.log
<AlmuHS>[557721.491] (WW) libpciaccess enabled in function VESAPciProbe
<AlmuHS>[557721.911] (WW) VESA(0): libpciaccess enabled in function VESAMapVidMem
<youpi>so PreInit was not getting called previously?
<youpi>this really look like the trigger is not inside the vesa driver, but xorg
<AlmuHS>maybe
<youpi>well, if none of the vesa functions that call libpciaccess are getting called, the libpciaccess issue can't be triggering inside the vesa driver
<AlmuHS>i've repeated the test using openbox http://sprunge.us/D0LdOV
<AlmuHS>openbox has crashed after a minutes
<AlmuHS>i had to kill Xorg with ctrl+alt+bksp
<AlmuHS>youpi: check the new log above
<AlmuHS>after kill xorg, kdb has crashed also
<youpi>well, ok, but I guess it's a different issue
<youpi>better chase only one issue at a time
<AlmuHS>yes. anyway, the openbox log maybe is simpler than Xfce log
<AlmuHS>the Xorg log using openbox, I refers
***Server sets mode: +nt
***Server sets mode: +nt
<AlmuHS>good night
<rekado_>hi there
***rekado_ is now known as rekado
<rekado>Hi there
<rekado>I’m running a Guixy Hurd VM and noticed that *often* it would fail to boot
<rekado>the last thing it prints is ‘start ext2fs: Hurd server bootstrap: ext2fs[device:hd0s1] exec’
<rekado>do you have any hints as to what I could do to figure out why this happens?
<janneke>oh, bad timing...youpi just left?
<janneke>rekado: i found there's a --debug switch: https://lists.gnu.org/archive/html/bug-hurd/2005-02/msg00043.html
<janneke>but it seems the ext2fs code suggests it needs to be compiled-in "#ifdef EXT2FS_DEBUG"
<janneke>lemme see
<janneke>hmm, "ext2fs.static --help" says we don't/didn't compile it in
<janneke>(not sure using --debug would tell us something...but yeah)
<janneke>ugh, it seems the source code needs to be patched to enable it
***pavlushka_ is now known as pavlushka
<gnu_srs>youpi: Is SIGCONT expected to work as expected?
<youpi>well, yes ?
<gnu_srs>OK, i just saw a test failure in src/libgo/go/os/signal/signal_test.go