This idea might have come up before, but it's hitting me like a brain wave and wants to be discussed. Let me know if you all think this should be taken over to f-devel-l. I'm mainly writing here because for this to happen, folks such as Will Woods need to be behind it.
Wondering if we could target having one or several live CDs/DVDs for the FC7 testX cycles.
There are multiple types of users who might run a test release but do not because they:
* Have only one system and are loathe to risk breaking it * Don't have enough experience or time to fix a system running rawhide or a test release * May have extra systems for testing, but not extra time to do a full install on each * Have older running hardware that is a PITA to install on, but easy to boot a CD on
If we combine this with e.g. a set of Dogtail scripts that users can add to, we make it easier for people to participate in testing.
For example, I could boot my system from a live DVD, load Dogtail and set of common scripts plus some scripts I make for myself to mimic my own workflow, then run those tests overnight while sleeping. In the morning, note any bugs etc., then reboot back into my working install of the current release of Fedora.
If we also make a weekly rawhide live DVD/CD available, we make it possible to include more of these folks throughout the development process.
The live CD concept allows us to create more than one set of packages to test. It is true that it would exclude other packages that are worth testing on the same system, but it opens the possibility for sub-projects to issue a special live CD just for testing e.g. Fedora Directory Server or the latest KDE packages.
- Karsten
On Saturday 11 November 2006 13:06, Karsten Wade wrote:
The live CD concept allows us to create more than one set of packages to test. It is true that it would exclude other packages that are worth testing on the same system, but it opens the possibility for sub-projects to issue a special live CD just for testing e.g. Fedora Directory Server or the latest KDE packages.
Live CD/DVDs would be invaluable in getting more testing of A) the liveCD creation tools, and B) the packages themselves. However it does not help to test anaconda and installation / upgrades. Its the installation/upgrades that had the most 'whoops' bugs in them and are the cause of most slips. We need more testing there, not less.
On Sat, Nov 11, 2006 at 01:14:13PM -0500, Jesse Keating wrote:
Live CD/DVDs would be invaluable in getting more testing of A) the liveCD creation tools, [...]
Are you talking about Kadischi here or are there other tools in use?
On Sat, Nov 11, 2006 at 01:47:36PM -0500, Jesse Keating wrote:
On Saturday 11 November 2006 13:43, Jos Vos wrote:
Are you talking about Kadischi here or are there other tools in use?
-EUNKNOWN
http://fedoraproject.org/wiki/Kadischi
What live CD creation tools were you talking about? And where can I get them? :-)
On Saturday 11 November 2006 13:51, Jos Vos wrote:
http://fedoraproject.org/wiki/Kadischi
What live CD creation tools were you talking about? And where can I get them?
I meant unknown if we'd use kadischi or some other tool. There are some concerns with the kadischi code base and how it works, which is why it isn't the 'official' Fedora live CD yet.
On Sat, Nov 11, 2006 at 01:56:28PM -0500, Jesse Keating wrote:
I meant unknown if we'd use kadischi or some other tool. There are some concerns with the kadischi code base and how it works, which is why it isn't the 'official' Fedora live CD yet.
OK, I subscribed myself to the live CD mailing list and I checked out Kadischi. I have to see how good it works, but I already saw way too much hardcoded lists in the code, that should be in a configuration file (that such a config file is just Python code is fine, but it should be separated from the code).
Furthermore, I don't immediately see why a live CD generator should (also) get a graphical user interface, but as I have the same opinion about a word processor :-), I won't further discuss this.
Anyway, I'll give it a try and start further discussions on the live CD list.
On Sat, 2006-11-11 at 13:14 -0500, Jesse Keating wrote:
Live CD/DVDs would be invaluable in getting more testing of A) the liveCD creation tools, and B) the packages themselves.
Yes, and combinations of packages that have a high degree of interoperability, such as GNOME or KDE.
However it does not help to test anaconda and installation / upgrades. Its the installation/upgrades that had the most 'whoops' bugs in them and are the cause of most slips. We need more testing there, not less.
The idea is to increase testing overall by including new groups that wouldn't test otherwise. Some of those are going to end up testing installation and upgrades, thereby increasing the testing on those.
Would some people stop testing installation/upgrades all-together? I suppose so, but it's all guesswork at this point. I guess that the numbers will offset each other and end up growing testing of rawhide and tests overall.
- Karsten
On Saturday 11 November 2006 18:27, Karsten Wade wrote:
The idea is to increase testing overall by including new groups that wouldn't test otherwise. Some of those are going to end up testing installation and upgrades, thereby increasing the testing on those.
Would some people stop testing installation/upgrades all-together? I suppose so, but it's all guesswork at this point. I guess that the numbers will offset each other and end up growing testing of rawhide and tests overall.
Indeed. I think its imperative that we soon decide on our official Live CD and start using it for testing purposes.
On Sat, Nov 11, 2006 at 13:14:13 -0500, Jesse Keating jkeating@redhat.com wrote:
On Saturday 11 November 2006 13:06, Karsten Wade wrote:
The live CD concept allows us to create more than one set of packages to test. It is true that it would exclude other packages that are worth testing on the same system, but it opens the possibility for sub-projects to issue a special live CD just for testing e.g. Fedora Directory Server or the latest KDE packages.
Live CD/DVDs would be invaluable in getting more testing of A) the liveCD creation tools, and B) the packages themselves. However it does not help to test anaconda and installation / upgrades. Its the installation/upgrades that had the most 'whoops' bugs in them and are the cause of most slips. We need more testing there, not less.
Telling people how to do their own respins might encourage more testing. Once I found a script to do it, I found that respins aren't all that hard to do. This allows you to test fixes since the last release without doing a lot of work. Personally, I am not likely to do much update testing, but I do keep spare partitions around that makes testing installs relatively safe and painless.
On Sunday 12 November 2006 13:55, Bruno Wolff III wrote:
Telling people how to do their own respins might encourage more testing. Once I found a script to do it, I found that respins aren't all that hard to do. This allows you to test fixes since the last release without doing a lot of work. Personally, I am not likely to do much update testing, but I do keep spare partitions around that makes testing installs relatively safe and painless.
Help out with pungi (:
http://linux.duke.edu/projects/pungi for the mercurial repo. No real "homepage" to speak of yet. I should fix that.
On Sun, Nov 12, 2006 at 16:53:40 -0500, Jesse Keating jkeating@redhat.com wrote:
On Sunday 12 November 2006 13:55, Bruno Wolff III wrote:
Telling people how to do their own respins might encourage more testing. Once I found a script to do it, I found that respins aren't all that hard to do. This allows you to test fixes since the last release without doing a lot of work. Personally, I am not likely to do much update testing, but I do keep spare partitions around that makes testing installs relatively safe and painless.
Help out with pungi (:
http://linux.duke.edu/projects/pungi for the mercurial repo. No real "homepage" to speak of yet. I should fix that.
For right now I have a usable script I got off the net. I'll try to keep pungi in mind(I bookmarked the page) and try it out eventually. But my free time is going to take a hit very shortly for about 4 months as I will be moving to a new house.
I spent a good chunk of this past weekend doing respins and fc6 installs and managed to find a bug (215231) in the combination of text based installs and software raid. I suspect that text based installs don't get tested too much any more.