PackMan - updating with local packages
David Pitt (3386) 1248 posts |
I have mixed feelings about PackMan, but have been constructively having another look. One issue I have occurs because I do some autobuilder builds, such GCC and its dependencies. Any build that I do from the latest autobuilder sources may have the same version number as previous builds, as found on PackMan, even if the source has been updated. As an example having installed Python from PackMan 20 Feb 12:34:27 000 00000000: Error from ‡|?: Version 4.0-Rel5-1 or later of package SharedLibs is already installed How should I get that later SOManager 3.02 installed? The version number cannot simply be incremented as the the next formally distributed update with up incremented version number could be missed. I can do something disgusting to Or the new version could be dropped on to the existing Is there a right answer to this? |
Steve Pampling (1551) 8170 posts |
Of no great use right now, possibly, but the BootUI test files that Julie Stamp is putting out seem to be headed the right direction for you and other developers in providing a route toward a setup that can be booted from a selection of drives. |
Chris Gransden (337) 1207 posts |
Normally you would wait until the next formal release (Rel6). You could just side load it in the mean time. It will get updated when Rel6 is released. When testing set up a separate autobuilder website. Basically run build-website and point PackMan at the package source file that is hosted on your your local web server. Also use a separate RISC OS machine with a freshly installed PackMan and no other package sources. You can also set up a Jenkins server to automate most of it. |
Paolo Fabio Zaino (28) 1882 posts |
@ David
Have you tried to just increase the “build id” of the RiscPkg package? the “-x” of the zip file name basically Every-time I do a build and package, even if it’s same version components, I always increase the build id number of the RiscPkg zip file name. Not sure if this helps with your specific problem, but this is how I test my RiscPkg installing them on top of the test system (in the early stages I was learning how to package this required even 3 or 4 tests for the same package). With that said, I totally understand your mixed feelings. I think PackMan needs some improvements, but I still think it’s the way forward. |
David Pitt (3386) 1248 posts |
I did not take this issue any further and reverted to not using PackMan. |