Phil Schaffner wrote:
Les Mikesell wrote on 04/11/2011 06:58 PM:
On 4/11/2011 5:32 PM, Ned Slider wrote:
...
It's laborious, it's repetitive, it's boring, sometimes it's time-consuming but it's really NOT difficult.
That depends on where and whether you can find the component(s) that were missing or the wrong version. But, it seems that if you have an after-the-build test, there might be a way to predict what you need to pass that test ahead of time - or at least to run all of the possible combinations in parallel if you really have to do trial-and-error.
Sounds a bit too much like AI to me.
Johnny addressed finding the components earlier, and they may not be discoverable.
http://lists.centos.org/pipermail/centos/2011-April/109631.html
If you have a way to do those predictive tests, in serial or parallel, I'm sure that would be a valuable contribution. The possible combinations quickly lead to a combinatorial explosion. Software regression testing is a science in its own right.
I think the best a community project can do on testing such packages is by flagging them for attention in a more open process with more testers.
Phil
I've already thought about this. When I try to recompile some package from Fedora for CentOS 5.x, I miss having package dependency database that is easy to follow.
I was thinking of creating following:
App that would parse primary.xml, filelists.xml and other.xml from each enabled repository and populate SQLite or MySQL database and would then allow to show tree-like dependencies for selected package. With all dependencies down to the last one, so if package1 depends on package2, package3 and, package4, and package2 depends on package6 and package7, all of those packages would be shown with respective version numbers.
The same XML format is for srpm repo folders, so we can have fast way of checking for dependencies.
Once you populate that database with srpms from all Fedora and RHEL versions (with all versions of every package), you could mix and match easier.
But I seam to be out of free time to start it, and my programming skills on Linux are currently stuck on very advanced bash.
Ljubomir