Re: About a mass bug report not based on Sid or Jessie.
2014-04-22 22:37 Santiago Vila:
Back in the early kfreebsd-* days, there was a server called
ftp.gnuab.org where every kind of hack was allowed in the source to
make packages build. Moreover, we could make NMUs at will without
having to ask the maintainer for permission, because they were for
the "unreleased" distribution. This allowed config.{guess,sub} and
similar bugs to be fixed quickly (at least in a temporary way).
Have things changed such at lot since then for new ports? How are they
made now?
I don't know if Wookey is doing something different with the help of the
mythical hardware, but I am using a quasi-mythical Qemu instead, poor man's
mythical hardware substitute.
But we are doing basically what you say for OpenRISC/or1k port, yes.
I do not understand your question very well, though. We have packages compiled
in our repositories with config.{guess,sub} and maybe disabling a
build-dependency here, or with empty documentation there, and it's fine. But
what after that?
Having 20k+ source packages and going through hundreds/thousands with outdated
config.*, and requesting them to update those files just for the set of ports
this year (when there are new ports every year), is not very scalable or
efficient -- even if maintainers acted so quickly as you did. Or upload source
NMUs -- I think that much bigger number now than the 200 that you mention in
kfreebsd time.
Basically, as discussed elsewhere in the thread, if we don't do something to
solve this in the packages themselves, we are condemned to repeat this work
forever, so if there's a way to put a stop to it, it would be great. Thanks to
all people participating in the thread for the suggestions.
On a personal note, I think that it is sad that something like autotools, which
was devised as a way to write portable code and so on, is one of the major
hindrances when it comes to create new Debian ports (at least, that's what I am
experiencing).
I completely agree with you, when you said in a previous message that this is a
serious design flaw.
Cheers.
--
Manuel
Reply to: