[ resend - now to the appropriate Fedora list - apologies ]
Hi everyone,
the olpc XS spin is hitting a problem installing on i586s (and that includes our own XO). The problem seems to be well known -- anaconda composes based on the arch of the build host rather than on the arch requested, as described in:
https://fedorahosted.org/revisor/wiki/AnacondaUpdates#TheUnabletoInstalloni5...
The revisor conf says "architecture=386", yet we are getting _only_ openssl i686 on the iso, which won't get installed on 586, so everything breaks on the (partially installed) machine.
I'm away from my buildbox today -- Jerry's been testing and reports that pungi-driven composes also put an openssl-i386 on the iso, while revisor-driven composes don't. So it sounds like the problem is perhaps in how revisor drives anaconda?
Any hints or ideas? It seems to be a well known problem...?
(the compose host is a Fedora-9 box, that follows updates.newkey)
cheers,
m -- martin.langhoff@gmail.com martin@laptop.org -- School Server Architect - ask interesting questions - don't get distracted with shiny stuff - working code first - http://wiki.laptop.org/go/User:Martinlanghoff
Hi Jeroen,
we have revisor on F-9 ignoring the requested arch and building the spin based on the host arch only - is this a known issue? The revisor-made spin we have doesn't work on i586 :-/
more details below...
On Sun, Feb 15, 2009 at 10:58 AM, Martin Langhoff martin.langhoff@gmail.com wrote:
the olpc XS spin is hitting a problem installing on i586s (and that includes our own XO). The problem seems to be well known -- anaconda composes based on the arch of the build host rather than on the arch requested, as described in:
https://fedorahosted.org/revisor/wiki/AnacondaUpdates#TheUnabletoInstalloni5...
The revisor conf says "architecture=386", yet we are getting _only_ openssl i686 on the iso, which won't get installed on 586, so everything breaks on the (partially installed) machine. I'm away from my buildbox today -- Jerry's been testing and reports that pungi-driven composes also put an openssl-i386 on the iso, while revisor-driven composes don't. So it sounds like the problem is perhaps in how revisor drives anaconda?
Any hints or ideas? It seems to be a well known problem...?
(the compose host is a Fedora-9 box, that follows updates.newkey)
cheers,
m
Martin Langhoff wrote:
Hi Jeroen,
we have revisor on F-9 ignoring the requested arch and building the spin based on the host arch only - is this a known issue? The revisor-made spin we have doesn't work on i586 :-/
more details below...
On Sun, Feb 15, 2009 at 10:58 AM, Martin Langhoff martin.langhoff@gmail.com wrote:
the olpc XS spin is hitting a problem installing on i586s (and that includes our own XO). The problem seems to be well known -- anaconda composes based on the arch of the build host rather than on the arch requested, as described in:
https://fedorahosted.org/revisor/wiki/AnacondaUpdates#TheUnabletoInstalloni5...
The revisor conf says "architecture=386", yet we are getting _only_ openssl i686 on the iso, which won't get installed on 586, so everything breaks on the (partially installed) machine. I'm away from my buildbox today -- Jerry's been testing and reports that pungi-driven composes also put an openssl-i386 on the iso, while revisor-driven composes don't. So it sounds like the problem is perhaps in how revisor drives anaconda?
Any hints or ideas? It seems to be a well known problem...?
(the compose host is a Fedora-9 box, that follows updates.newkey)
Hi Martin,
Since your first mail, I've tried to reproduce this. I have no problems producing installation media with both openssl.i386 and openssl.i686 in the RPM payload (Packages/ directory) - with either respin mode or without, which I was thinking may have been the issue. This however is using a fresh git clone off of master, not the released version of Revisor on Fedora 9.
Are you composing Live media by any chance? I can see how there is no .i386 version of openssl installed.
Kind regards,
Jeroen van Meeuwen -kanarip
On Mon, Feb 16, 2009 at 2:52 PM, Jeroen van Meeuwen kanarip@kanarip.com wrote:
Since your first mail, I've tried to reproduce this. I have no problems producing installation media with both openssl.i386 and openssl.i686 in the RPM payload (Packages/ directory) - with either respin mode or without, which I was thinking may have been the issue. This however is using a fresh git clone off of master, not the released version of Revisor on Fedora 9.
We are not using respin mode, we're creating it from scratch...
Are you composing Live media by any chance? I can see how there is no .i386 version of openssl installed.
No. Just a single disk install iso...
cheers,
m
Martin Langhoff wrote:
On Mon, Feb 16, 2009 at 2:52 PM, Jeroen van Meeuwen kanarip@kanarip.com wrote:
Since your first mail, I've tried to reproduce this. I have no problems producing installation media with both openssl.i386 and openssl.i686 in the RPM payload (Packages/ directory) - with either respin mode or without, which I was thinking may have been the issue. This however is using a fresh git clone off of master, not the released version of Revisor on Fedora 9.
We are not using respin mode, we're creating it from scratch...
Like I said either with or without respin mode, both openssl.i386 and openssl.i686 are pulled into the RPM payload.
Can you maybe send me a log file with both Revisor as well as YUM set to debuglevel 9?
For Revisor, that's --debug 9, for YUM that's debuglevel= in the appropriate configuration file in /etc/revisor/conf.d/
Thanks,
Kind regards,
Jeroen van Meeuwen -kanarip
On Tue, Feb 17, 2009 at 3:17 AM, Jeroen van Meeuwen kanarip@kanarip.com wrote:
Can you maybe send me a log file with both Revisor as well as YUM set to debuglevel 9?
Hi Jeroen,
last week you mentioned the logs made sense... any news on this track? Did this evolve into a different thread subject and I missed it...?
thanks for your help!
cheers,
m
Martin Langhoff wrote:
On Tue, Feb 17, 2009 at 3:17 AM, Jeroen van Meeuwen kanarip@kanarip.com wrote:
Can you maybe send me a log file with both Revisor as well as YUM set to debuglevel 9?
Hi Jeroen,
last week you mentioned the logs made sense... any news on this track? Did this evolve into a different thread subject and I missed it...?
Hi Martin,
did I forget to reply with a commit code? Sorry!
I did this commit http://git.fedorahosted.org/git/?p=revisor;a=commitdiff;h=c76d857046e931121a...
to prevent F9 and above to end up with no allarch packages being selected. Could you test this and see if it helps?
Kind regards,
Jeroen van Meeuwen -kanarip
On Tue, Feb 24, 2009 at 1:15 AM, Jeroen van Meeuwen kanarip@kanarip.com wrote:
did I forget to reply with a commit code? Sorry!
I did this commit http://git.fedorahosted.org/git/?p=revisor;a=commitdiff;h=c76d857046e931121a...
to prevent F9 and above to end up with no allarch packages being selected. Could you test this and see if it helps?
Works for Jerry and for me. Fantastic - thanks!
Are you planning on updating the F-9 or F-11 packages with it?
I built a package based on the F-9 branch branch of your git repo, and the version numbers don't like up with the latest rpm on F-9 (2.1.1-5 vs 2.1.1-7), and I think I saw a minor regresion (the ISO is called Fedora instead of OLPC School Server, .discinfo is correct however) so I'm pretty sure the RPMs in F-9 don't come from there.
I was hoping to custom-build rpms based on the exact latest F-9 src plus with this cherry-picked bugfix...
cheers,
m
On Tue, Feb 24, 2009 at 3:12 PM, Martin Langhoff martin.langhoff@gmail.com wrote:
Are you planning on updating the F-9 or F-11 packages with it?
Hi Jeroen,
- is there a revision coming on the F-9 or F10 branches? - can you check whether the F-9 package (2.1.1-7) matches what's in git? From what I can see, it doesn't, so I can't build a local rvisor pkg I can trust :-/ ...
cheers,
m
Martin Langhoff wrote:
On Tue, Feb 24, 2009 at 3:12 PM, Martin Langhoff martin.langhoff@gmail.com wrote:
Are you planning on updating the F-9 or F-11 packages with it?
Hi Jeroen,
- is there a revision coming on the F-9 or F10 branches?
I'm planning a release for F-10.
- can you check whether the F-9 package (2.1.1-7) matches what's in
git? From what I can see, it doesn't, so I can't build a local rvisor pkg I can trust :-/ ...
The packages releases are always behind the GIT repo, because the packages get created from what is in the GIT repo (at some point).
If I gave you some 2.1.4 revisor packages to test on Fedora 9 (I'm not sure they work but there have been no major changes), are you able to test them?
Kind regards,
Jeroen
On Thu, Feb 26, 2009 at 9:46 PM, Jeroen van Meeuwen kanarip@kanarip.com wrote:
If I gave you some 2.1.4 revisor packages to test on Fedora 9 (I'm not sure they work but there have been no major changes), are you able to test them?
Yes. My only 'compat' concern is with anaconda. If they expect anaconda smarts that aren't there...
m
Martin Langhoff wrote:
On Thu, Feb 26, 2009 at 9:46 PM, Jeroen van Meeuwen kanarip@kanarip.com wrote:
If I gave you some 2.1.4 revisor packages to test on Fedora 9 (I'm not sure they work but there have been no major changes), are you able to test them?
Yes. My only 'compat' concern is with anaconda. If they expect anaconda smarts that aren't there...
Well, Revisor (by means of the version-specific buildinstall scripts in /usr/lib/revisor/scripts/) on Fedora 10 is able to compose Fedora 9, and only changes in things -like squashfs major version mismatches (like right now between a Fedora 9+ and an EL-5 station) may cause issues. The Revisor version itself... shouldn't cause all that many issues.
If you want a quick try you can do:
autoreconf -v && ./configure && make rpm in the source tree.
Kind regards,
Jeroen van Meeuwen -kanarip
buildsys@lists.fedoraproject.org