patches in the 2.0 series...
Ryan C. Gordon
icculus at clutteredmind.org
Sun Feb 16 11:07:36 EST 2003
I tend to bitch about loki_setup approximately 24 hours after any given
ut2003 patch, I know, but here's what I'm thinking currently.
We need to change the way we do patches with setup 2.0.
First, they should _not_ be self-extracting. This has caused no end of
trouble and suffering due to glibc issues, etc on loki_patch. Plus, if
something external handles extracting, we can guarantee that a) md5sum is
available (as part of the extraction program, not an external utility) and
b) bzip2 decompression is available (we're more or less forced to have
bigger, gzip'd patches at the moment...EVERY time I put out a tarball
some server admin goes "what do I do with a .tar.bz2 file?!" and some
systems STILL don't have a standard bunzip2 binary on them).
Second, the patch program should not be included with the patch. When the
patch breaks because of this, the only answer is "tough luck, kid"...or,
"here, rip the patch apart and add these files and rerun it and clean up
afterwards...wasn't this so user friendly?" Also, including the same
binary with every patch makes the thing unnecessarily bigger...moreso if
there are multiple architectures that the patch applies to.
loki_update should do the patching itself once it downloads the patch
file...or loki_patch should be part of loki_update. Patches should be data
that the tools work on, not packages themselves.
More information about the Lokisetup