PPM: post uninstall script

Forum member shinobar reported that the Puppy Package Manager does not support the 'puninstall.sh' script. Yes, it is a complete rewrite, compared with PETget in 4.21 and earlier, and I had not yet implemented that feature.

The PET package specification states that two scripts are allowed, a post-install script 'pinstall.sh' and a post-uninstall script 'puninstall.sh'.

Ok, I have edited /usr/local/petget/installpkg.sh and /usr/local/petget/removepreview.sh, now supporting the post-uninstall script.

Note, at package installation 'puninstall.sh' gets moved from '/' and renamed to /root/.packages/<package-name>.remove

The idea of a post-uninstall script is to do some essential cleanup that simply removing the packages files has not achieved. However, it is wise to look through the 'removepreview.sh' script to see how it uninstalls a package, as quite a lot of post-install cleanup is automatically done -- like making sure that a file on a lower Unionfs layer that was visible on-top prior to the package being installed, is once again visible after the package is uninstalled.

Posted on 18 Aug 2009, 15:37


Posted on 18 Aug 2009, 23:15 by Sit Heel Speak
Question on new 4.23 petget
Thanks--lack of puninstall.sh has been stopping risk-averse moi from keeping current with my bci pets.

Will the rewritten versions be useable in 4.22 and in earlier versions of 4.23?

A related issue I've noticed: beginning in 4.22, is that PPM/petget will error-out if the PET is not downloaded the usual way into /root but rather resides in a subdir whose dirname has a space in it.

Not a problem for users, but a profanity-inducer for anyone trying to develop multiple versions of the same PET and keep them in separate, explicitly-labelled subdirs.

Posted on 19 Aug 2009, 8:45 by BarryK
I don't know what you are asking here:

Will the rewritten versions be useable in 4.22 and in earlier versions of 4.23?

I'll look at the petget script, see if I can fix the problem with spaces.

Posted on 19 Aug 2009, 13:03 by Sit Heel Speak
petget clarification
>>I don't know what you are asking here

Well, for a specific example, I'm wondering if I could supply the revised petget script within a bci PET to Beem for his upup-466, so he can uninstall my handiwork if he doesn't like what it does to his fonts!

Posted on 19 Aug 2009, 18:30 by BarryK
petget scripts inside .pet
It could be done, but not a very good idea as those scripts would then install on top of the official ones. The official scripts will continue to evolve.

It is better if developers/users upgrade to the latest Woof build system.

Note, I plan to do another Upup build soon.

Posted on 20 Aug 2009, 2:03 by Sit Heel Speak
petget substitution / bci rethink
Acknowledged. I will play-it-by-ear on whether to supply petget with any particular bci pet.

Two Pup(py) versions can have the same kernel, gcc, libc.so.6 and still the bci pet which beautifies the one will uglify the other. So, I surmise that gtk+ and related libs, pango+cairo, libXft, and the browser are compiled from-Puppy-version-to-version with different-enough switches and/or evolving dependencies to require I compile bci-libfreetype on the specific Pup(py) on which it will run.

Some must lag back, e.g. Beem must use upup-466, not 476, because his Medion and 2nd sata drive don't work with 476. Such omissions inevitably afflict the newest, Beem's is but one typical case of many, so, as I say, I'll play it by ear.

Something else has digeridoo'ed over my kazoo: the whole idea of the bci pet has been called into question, at least for some Pup(pie)s, by what I've learned from iguleder and Beem. iguleder posted a very simple few-lines edit which 4.12, 4.21, and upup-466, and perhaps all Pup(pie)s, benefit from practically as much as from my 660kb bci-hinting pets. It's just a few-lines edit to /etc/fonts/local.conf, and/or creation of a /root/.fonts.conf. (at least, it works practically as well as my bci pets on Puppy's default DejaVu Sans font; haven't tested it yet on other fonts). Heh...I'm a little embarrassed at the reactionary way I greeted this shockingly wonderful revelation. Well, at least it was a revelation to me...

upup-466 to a small extent, and official Puppy 4.12-regular to a huge extent, look positively awful if I apply bci-hinting.

Posted on 20 Aug 2009, 6:28 by disciple
need bci
IMO if you want really good fonts you still do need either bitmap fonts or bci-hinting + some well hinted fonts like the MS core web fonts.

Posted on 20 Aug 2009, 6:30 by disciple
Although I do think all the font substitution you do is complicated and unnecessary :)
I just use Arial for everything :)

Posted on 20 Aug 2009, 7:44 by Sit Heel Speak
Ode to Arial (and WhoDo)
*lol* yeah, Arial is the main reason I chose to figure-out bci-hinting in the first place. Arial bold 10 is the GTK+ theme font in the screenshot (it's IceWM with the late lamented and sorely missed WhoDo's Vista Black IV theme from 2.15CE EZPup, of course, tinkered-with incompetently ;-} ) O WhoDo, please, return and save me from myself!!! Verdana bold 10 is what I label the icons with, in rox options Pinboard Custom Font. Whenever I feel like rearranging the furniture I usually go for Lucida Sans Unicode.