mindi syslinux problem

[victor gattegno]

Mindi uses "syslinux -s" to make a media bootable, it was working well until Ubuntu 10.04. But, on Ubuntu 10.10 and on Ubuntu 11.04 the version of syslinux is 4.02.

Problem: with syslinux v 4.02, the -s option returns an error, then an USB backup hard disk made by mondoarchive is not bootable.

The error can be found in mondoarchive.log :

syslinux: invalid option -- 's'
syslinux: Unknown option: -s
...
Failed to create USB image.
Boot and data disk images were created.

And in the root of the USB backup hard disk made by mondoarchive, ldlinux.sys is missing.

A solution is to modify /usr/sbin/mindi by replacing "syslinux -s" by "syslinux --stupid", and to run mondoarchive after that.

Another solution is to run, after the mondoarchive command (if by example the external USB hard disk is /dev/sdb): # syslinux --stupid /dev/sdb1

That problem is hopefully corrected in syslinux v 4.03, but Ubuntu 10.10 and 11.04 don't propose the update.

On Ubuntu 11.10 the version of syslinux is 4.04, so there is no problem with that version.

The shell-script /usr/sbin/mindi should be modified to avoid problem with syslinux version 4.0.2, that version of syslinux can be included in some other GNU/Linux distributions.

[Bruno Cornec]

Well, for the moment, I mainly see Ubuntu affected. And I tend to consider that as a bug in the distribution, not in mindi which does the right thing. To be more precise it's a bug in syslinux upstream, but upstream fixed it ! So now the distro can backport the fix into their version easily. So maybe the best would be to open a bug report wrt Ubuntu on syslinux and have them fix it.

I have e.g. a bug at Red Hat for acl that I declared upstream and that should be at last fixed with RHEL 5.8, but I've not done workaround for that in the code, as I really consider it's their fault to not have included the upstream fix.

So it's good this Bug report exist to hopefully help the one using google to find info on that issue. what we could add to it is a patch for them to apply on their delivery, and I could even consider integrating it in the Ubuntu packages, but not in upstream as mindi usage is just correct.

[Bruno Cornec]

Ok, so I now have made patches to support this for the related Ubuntu versions so that it will always been generated correctly in the future. This required to add support for pathces to deb based distros, which project-builder.org didn't had up to now. So example at ​ftp://ftp.mondorescue.org/test/ubuntu/10.10/mindi_2.1.0-0.20111127025447_amd64.deb includes that fix built with rev 1367 of project-builder. Will be fixed for next 3.0 version of mondo, when built with the next version of pb and the VMs updated.

[victor gattegno]

Same problem with CentOS 6 and RHEL 6

Someone just got that problem with CentOS 6.2 too.

Maybe a general solution could be to execute twice the command in /usr/sbin/mindi, so, if the first one fails because of syslinux 4.02, the second one succeeds :

syslinux -s /dev/sdb1
syslinux --stupid /dev/sdb1

Another solution could be to test the version of syslinux, and, if it is v 4.02, to use "--stupid" option instead of "-s" option.

Example : on my RHEL 6.2 (GOOD LORD I get also a 4.02 syslinux !!)

# syslinux -v 
syslinux 4.02  Copyright 1994-2010 H. Peter Anvin et al

[victor gattegno]

Correction : I use RHEL 6.1, but maybe on RHEL 6.2 syslinux v 4.02 is still used...

[victor gattegno]

Note : I checked mindi source (MINDI_REV=2930) and I don't see the fix for Ubuntu neither (I don't see any "syslinux --stupid").

[Bruno Cornec]

Replying to vicgat:

Note : I checked mindi source (MINDI_REV=2930) and I don't see the fix for Ubuntu neither (I don't see any "syslinux --stupid").

This is normal as the fix is done through a patch applied at pckage build time. Source wasn't touched. Will check version then for solving it correctly.

[Bruno Cornec]

Fixed differently (in the code) in rev [2939]