[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#778627: live-build: syslinux mbr.bin file not found (again)



On 22 February 2015 at 15:23, jnqnfe <jnqnfe@gmail.com> wrote:
> On Sun, 2015-02-22 at 13:59 +1100, David wrote:
>> On 22 February 2015 at 10:04, Michal Suchanek <hramrach@gmail.com> wrote:
>> > On 21 February 2015 at 19:29, jnqnfe <jnqnfe@gmail.com> wrote:
>> >> On Tue, 2015-02-17 at 17:30 +0100, Michal Suchanek wrote:
>> >>> it seems mbr.bin has moved to a subdirectory which causes hdd build
>> >>> failure in Jessie.
>> >>>
>> >>> Attaching a patch.
>> >>
>> >> Sorry, where has this change occurred? Looking at the list of files
>> >> installed, e.g. [1] the file seems to be in exactly the same place as
>> >> live-build is currently looking... Or is this info out of date with
>> >> respect to what the package is installing and where??
>> >>
>> >> [1] https://packages.debian.org/sid/amd64/syslinux/filelist
>> >>
>> >
>> > Actually, I would like to know too. The build was clearly failing at
>> > that point saying there is no such file and I had the mbr subdirectory
>> > in chroot but now the chroot does not have it.
>>
>> In wheezy [2] the syslinux-common package installs:
>>   /usr/lib/syslinux/mbr.bin
>>
>> In jessie [3] this is unchanged only for [ia64, sparc, s390]
>> architectures, apparently for all other architectures it has been
>> changed to install to a sub-directory named "mbr":
>>   /usr/lib/syslinux/mbr/mbr.bin
>>
>> And this breaks the references to mbr.bin in the binary_hdd script
>> for these architectures.
>>
>> [2] https://packages.debian.org/search?searchon=contents&keywords=mbr.bin&mode=exactfilename&suite=stable&arch=any
>>
>> [3] https://packages.debian.org/search?searchon=contents&keywords=mbr.bin&mode=exactfilename&suite=testing&arch=any
>
> According to your [3], it is in the syslinux-common package where this
> change has occurred (if that wasn't how it already was previously). The
> 'syslinux' and 'extlinux' packages, which also carry a copy do not have
> it in an 'mbr' subdirectory, thus they have it in exactly the location
> the script is looking for it.
>
> The binary_hdd script uses either the syslinux or extlinux package,
> depending on what filesystem format it is going to use. It does not use
> the syslinux-common package. However the syslinux-common package may
> have been installed by another script in live-build, from which the
> 'mbr' directory was not deleted on removal at the end of that script,
> thus explaining its presence.
>
> I am not denying outright there's a problem, clearly the build failed,
> otherwise this discussion would not have begun. Perhaps when I can find
> a little time soon I'll run a HDD based build to check, but just on the
> face of it, looking at the info in the package system, I do not see why
> binary_hdd should need to look in in an 'mbr' sub-folder, and why it
> would fail from not being able to find the mbr.bin file.
>
> @Michal have you re-tried the build process since? (note, it's sometimes
> a good idea to run lb --clean before hand if a failure occurred). I.e.,
> is this problem repeatable, or may it just have been a fluke, for
> example, maybe apt failed to download the syslinux/extlinux package and
> yet the build carried on anyway, then understandably failed to find the
> file. (Just a wild guess at a possible explanation).
>
> If it's not repeatable, and may have just been an odd fluke, we should
> close the bug.

Some months ago I did encounter repeatable live-build failures due to
this mbr.bin location and I temporarily worked around it by patching
my binary_hdd script.

It might be important to mention that (because my target hardware bios
fails when given usb flashdrive iso-hybrid images) it is necessary for
me to build with '--binary-images hdd'. So I suggest testing with that
setting.

Everything I was doing was scripted, from clean starts, and tracked
with git, so I would not categorise this bug as an "odd fluke" and I
should be able to reproduce it, but I apologise that I do not have any
free time at the moment to revisit this.

Thanks for all your other work on live-build!


Reply to: