| Ady2 | I downloaded lubuntu-15.04-alternate-i386.iso. Inside it, I see 'install/netboot/ubuntu-installer/i386/pxelinux.cfg', but it is empty. I also see 'install/netboot/ubuntu-installer/i386/pxelinux.0' (not under the 'pxelinux.cfg' subdir) but there is no ldlinux.c32 in the same directory (so pxelinux.0 would fail to find it). Is there any place where I can find info about this (e.g. relevant package, or how to use | 08:28 |
|---|---|---|
| Ady2 | / what's the goal of this directory and its subdirs.)? I cannot make sense of these subdirs. Should they be included in this iso image at all? What for? Is this some kind of bug? | 08:28 |
| Kamilion | Not sure, Ady2 | 11:26 |
| Kamilion | I've had a number of small issues with mismash files on some of the ISOs | 11:26 |
| Kamilion | for example, mboot.c32 is also AWOL | 11:27 |
| Kamilion | or was, at least | 11:27 |
| Ady2 | Kamilion: I am investigating it. I have a hunch this is sth related to debianinstall or related package. I would like to know where are these files and directories coming from. I don't have a problem of mismatch versions, just that there should be one additional file in there (which is located in a different subdirectory). | 11:28 |
| Kamilion | by mishmash I mean that some files are present and others are missing. Seems mostly syslinux/isolinux/pxelinux modules. | 11:29 |
| Kamilion | I couldn't be tossed to figure it out, so I just threw mboot.c32 in there myself. (i needed it to load xen) | 11:30 |
| Ady2 | I also don't know why the files are not under 'pxelinux.cfg/', but first I guess I should find where are these files and structure originated. Yes, they are from Syslinux, but the structure is not from Syslinux upstream. Someone else (debianinstaller?) puts them there. | 11:31 |
| Ady2 | And finally, I would like to know why they are included in the iso. How they are supposed to be used, under which scenario, why we need additional syslinux files, and additional kernel and initrd. | 11:32 |
| Kamilion | last time I looked into the matter I was lead to https://bazaar.launchpad.net/~ubuntu-cdimage/ubuntu-cdimage/mainline/view/head:/bin/cron.daily-live | 11:32 |
| Kamilion | in your case it would have been https://bazaar.launchpad.net/~ubuntu-cdimage/ubuntu-cdimage/mainline/view/head:/bin/cron.daily | 11:34 |
| Kamilion | this is apparently the mechanism that actually makes the ISOs | 11:35 |
| Kamilion | I'm still trying to grasp how everything works for a .seed to end up as a built ISO set | 11:44 |
| Kamilion | I hit a brick wall at https://bazaar.launchpad.net/~ubuntu-cdimage/ubuntu-cdimage/mainline/view/head:/lib/cdimage/livefs.py#L166 | 11:49 |
| Kamilion | it looks like that's the point where it goes off and runs the package builds needed to end up on disc | 11:52 |
| Kamilion | but deeper in there is where it grabs the build artifacts, including all of the files. https://bazaar.launchpad.net/~ubuntu-cdimage/ubuntu-cdimage/mainline/view/head:/lib/cdimage/livefs.py#L706 | 11:56 |
| Kamilion | s/all of the files/all of the boot files/i | 11:57 |
| Ady2 | Considering http://cdimage.debian.org/debian/dists/jessie/main/installer-i386/current/images/netboot/ this has to be originated somewhere in debian too, perhaps in the d-i-netboot-assistant package. But then it goes to ubuntu and then to lubuntu too. So the Q is, is there someone here that can give some clue? | 12:44 |
| Ady2 | The equivalent to what it is seen inside the iso is http://cdimage.debian.org/debian/dists/jessie/main/installer-i386/current/images/netboot/debian-installer/i386/ | 12:45 |
| Kamilion | yeah, someone'll show up with more knowledge of history than I | 12:52 |
| Kamilion | So do your best to explain the situation now and someone'll read the scrollback. | 12:53 |
| Kamilion | at the very least, someone will tell you which package the bug report should get filed on. | 12:55 |
| Ady2 | I don't have a situation. I am trying to understand why all these subdirectories and files are included in the iso, and I know Syslinux enough so to understand that, for whichever goal these have been included, they still need ldlinux.c32 together with pxelinux.0, or it will fail in whichever situation these files and directory structure could be needed. | 12:57 |
| Kamilion | my guess would be filing a bug against syslinux or d-i. I know the relevant files exist in the built package. Why they don't arrive on the media? *shrug* Someone here probably knows when they wake and read the scrollback. | 13:02 |
| Kamilion | only certain files are copied, according to some rule during ISO generation. Where that lives; I'm still trying to find out. | 13:03 |
| Kamilion | If you happen to find out -- let me know, so I can request mboot.c32 too | 13:03 |
| Kamilion | I've got to go get some sleep; but I didn't want you to leave thinking you were ignored. | 13:04 |
| Ady2 | TY. FWIW, the ldlinux.c32 is there, but in the wrong subdir. | 13:04 |
| Kamilion | well, either way, it's not something that can be fixed; the release has came and gone and the ISOs are already mirrored worldwide | 13:05 |
| Kamilion | so it will have to be fixed for 15.10 | 13:05 |
| Ady2 | The problem is also described at https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=750586#90 , expect that the bug is not upstream Syslinux, but in Debian, in some package related to DebianInstaller (and my guess is that such package is netboot-assistant). It doesn't answer my doubt though, why are these files and directory structure included in the iso (I mean, what for)? | 13:21 |
| ubot93 | Debian bug 750586 in debian-installer "syslinux-common: Boot fails. Failed to load ldlinux.c32. File must be in /. Upstream bug" [Important,Open] | 13:21 |
| Kamilion | ... Huh. | 13:27 |
| Kamilion | yeah, THAT issue sounds like an upstream bug in syslinux (re: pxelinux.0 not respecting ldlinux.c32's path) | 13:28 |
| Kamilion | something has changed recently | 13:29 |
| Kamilion | ~march/april | 13:29 |
| Ady2 | I repeat, the bug is not in Syslinux itself, but in Debian. The fact that users report something being upstream because they happen to find that a prior version works doesn't mean their conclusions are correct. | 13:32 |
| Kamilion | In general, regressions are A Bad Thing, in my opinion. | 13:36 |
| Ady2 | The fact that the reporter in msg #90 just copies ldlinux.c32 to the "right" location (as per Syslinux expects it) and it works actually shows that the problem is the location in which the Debian package is putting the files. Whether it is one file that needs relocation (ldlinux.c32), or the whole structure is inadequate, I don't know). The package (netboot-assistant) is not maintained (correctly). | 13:36 |
| Ady2 | This is not exactly a regression. It just needs further modifications in addition to whatever they did during the last months. | 13:38 |
| Kamilion | I understand your sentiment. | 13:38 |
| Kamilion | In general from the deployments I've seen, most PXE setups tend to be site-specific, and customized to a high degree. | 13:40 |
| Kamilion | A great many of the ones I've recently interacted with have moved over to using iPXE and loading files from a protocol newer than TFTP, such as HTTP or HTTPS. | 13:42 |
| Kamilion | and the last time I checked, that process currently doesn't need pxelinux.0 | 13:43 |
| Kamilion | so my observation is that it may be not be maintained (correctly) due to people moving on to a more modern solution. | 13:44 |
| Kamilion | leading to things like pxelinux.0 not getting as much testing as it once did; where people would cry out when something changed for the worse. | 13:44 |
| Ady2 | Sure, I understand. Well, v.5 added lpxelinux.0 already, but that's not my point here. For whichever reason, this directory structure is included in the iso, and I really would like to know what's its purpose. | 13:46 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!