jrb wrote: Fri Mar 03, 2023 3:23 pm
Marv wrote: Fri Mar 03, 2023 2:54 pm
Good -and don't forget to sleep- I briefly looked at what it would make to set a symlinked pup flag variable and use that to trigger my 'catcher' and possibly put your real fixes in a function and trigger them with it too, but each of the package flavors are just enough different in how they are unpacked so that was both unwieldy and wound up saving little space even though it would keep the installpkg.sh more universal. You probably already thought that through.
Thanks, Jim
Thought about but no solution yet. Would like some test that could make this compatible with all the puppybuilds in WoofCE.
P.S. Its not forgetting to sleep, its forgetting how to sleep. 
I'll chew on that 'universal' a bit more. Less urgency now and maybe the third time is the charm.
@jrb Ya, caught that in 0.2, one of my bad pets only had a /bin. I was working on a 'universal' installpkg.sh. so I just fixed it in the library_fixer function and carried on. The first and probably pretty raw go-round is attached, false gz as usual, For jrb only for now but forum won't let me attach to PM any more?
Tested only with pets so far, I will do more but attached just to let you know I am chewing on it.
A couple more tests, using libcap2-bin_2.44, the problem library from the gthumbs install. Installed both from PPM and from a downloaded deb. Ok in both, the problem files were relocated from /sbin to /usr/sbin. I'll put it in a non-symlinked pup next, probably F96-CE_2 and check installs for pets and debs there, both local and PPM.
OK, hopped over to F96-CE_2 with the library_fixer installpkg.sh on board. The same set of pets, debs, and PPM install as above. All successful with the libraries and executables all showing up in the 'traditional' locations.
Seems to be working, Cut2 (attached) passes pkg types to the library_fixer function a little more neatly. No other changes.
Renamed control variable SYMLINKEDLIBS to SYMLINKEDPATHS, set to false if no /symlinks found,
renamed library_fixer to path_fixer, inverted logic where called so it tests for "=true"
No change at all in function. attached as cut3
I'm pulling the attempt to extend it to rpms etc. (cut4) Both messy and failed tests. Cut3a is my current best attempt.
Jim