Ben Hines <bhines@...> wrote:
> On Sunday, December 22, 2002, at 11:48 PM, David R. Morrison wrote:
> > I did this because it has become clear that quite a bit of testing
> > still
> > needs to be done, and some existing Fink packages will need to be
> > modified,
> For example, which packages still don't work? gd? You fixed it.
Anytime we want to make a fundamental change in the Fink build environment,
somebody needs to verify that every existing package (or at least every
stable package) continues to work when the change is made. Nobody has
done this for the -isystem/LIBRARY_PATH/DYLD_LIBRARY_PATH change.
I discovered quite by accident recently that my fvwm2 was not building
properly when the -isystem change was implemented. I fixed it. But
how many other problems are hiding out there? If you introduce a
problem in a stable package which was not present before, you have
weakened the stability of the system.
> > TOT on the main branch, so that we can release fink-0.11.2 quite soon.
> Well, unlikely.. current cvs also has some rather severe splitoff bugs
> (they don't work), i am pretty sure caused by RangerRick's recent
> patch. I was going to suggest releasing current fink (with isystem)
> once those are fixed. But its very much not releasable now and those
> bugs are worse than the isystem changes (which can be fixed by easy
> settings in the info file.)
Well, I hadn't heard of problems with the other patches to fink, and in
fact Justin has been lobbying pretty hard to get some repaired version
released that handles the dependency problems.
When you say that the isystem changes can be easily fixed, that is in
some sense true. We could go through and add a few lines such as
to every package (with one caveat, mentioned below), which would cause
the package to follow the old default behavior under the new system.
It would then be up to individual maintainers to modify this if they
The caveat is, if the package actually sets CPPFLAGS or LDFLAGS
then we have to add to the value rather than replace is... makes scripting
this operation slightly more complicated, but still possible.
Anyway, the main point of this message is that there is definitely still
some nontrivial work to do before the -isystem stuff can be released.
If there is also work to do about other patches, I'll be happy to back
those out for a while, also, so that we can get an up-to-date fink out