Hugo Osvaldo Barrera wrote:
As maintainer of python2-pyside shouldn't I have gotten a notification about this? At least an automated one from the AUR?
Also, since the PKGBUILD states "replaces=(python2-pyside...", shouldn't searching the AUR for python2-pyside yield python-pyside as one of the results?
The AUR does not fully parse PKGBUILDs because that would require a tool that can fully parse Bash without executing it, which no one has written yet (to my knowledge). As such, the current approach fudges it with simple regexes or whatever that only extract a subset of data. I do not think that includes "replaces". The AUR also lacks true support for split packages (again, because there is no full Bash parser*), which is why the python2-pyside split package is not detected.
Finally, why does this merge make sense? It merges two *different* libraries into one; I maintained python2-pyside, but have no interest in installing the python3 version, so why is the original package deleted?
A split package builds 2 or more packages. You may optionally install any of them as needed. It makes sense here to have a split package because both the Python 2 and Python 3 versions are built from the same source. The only argument against having a split package in this case is the lack of AUR support, which makes the package a little harder to find and which requires some ugly hacks to the PKGBUILD (e.g. "true &&" to trick the parser). Regards, Xyne * This is the price we pay for having all metadata trapped in a general purpose, quirky scripting language that can only be evaluated by executing arbitrary code.