Quote:
Originally Posted by frmald
That's wonderful and I appreciate that, but perhaps you might consider packaging all useful resources (like the desktop file?) in the "isolated install", to make repackaging easier.
|
Quote:
Originally Posted by frmald
Because putting a normal desktop file in the install directory, or considering independent packaging efforts was... too ideologically disturbing?
|
There's nothing remotely extraordinary about a program that generates desktop files in the Makefile or similar. Which is precisely what `python setup.py install` is.
I don't much care about the binary installer, but as I mentioned above it's not hard to just pass --root to it (and play Freedesktop XDG-utils games for overriding their default target). So I don't see how this is ideologically a battle between ideologies. Given the files are dynamically generated from up to date info, I don't see much point in shipping them again and having postinstall merely copy them. It sounds more complex, essentially busywork.
Quote:
Originally Posted by frmald
Anyway, who am I kidding, calibre even uses BINARY files, instead of scripts, to set up paths. Hence, ebook-viewer vs bin/ebook-viewer. Wonderfully non-discoverable and opaque. I *suppose* it's just a matter of setting up LD_LIBRARY_PATH, but who knows, there's stuff both in lib/ and libexec/.
|
The from source version uses python scripts.
The binary version is based on
https://github.com/kovidgoyal/calibr...nux/launcher.c
I'm not sure why this is especially evil, to use C launchers instead of shellscripts. Why is it supposed to be discoverable? Is a C program that *isn't* a launcher also supposed to be discoverable without disassembling?