[Pythonmac-SIG] Re: bundlebuilder bug?

[Russell E. Owen]

In article <7CA5094E-4DEF-11D8-824A-000A95686CD8 at redivi.com>,
 Bob Ippolito <bob at redivi.com> wrote:

> On Jan 23, 2004, at 3:49 PM, Russell E. Owen wrote:
> 
> > I just submitted this to sourceforge, but thought I'd see if anybody 
> > had
> > any suggested workarounds. I'm also unclear if I should be reporting a
> > 2nd bug as well.
> 
> Next time it'd be cool to get a URL for it in the bug tracker, only 
> Jack gets email notifications about this stuff, and it's sort of a pain 
> to dig around in sourceforge's bug tracker.

Sorry about that. I'll try to remember for next time.

> > The bundlebuilder command-line argument --lib is supposed to be able to
> > include frameworks in the resulting Mac application package.
> > Unfortunately, the aliases in the resulting framework copy are all
> > broken....

I suspect you're right, since things now seem to be running OK.

I went on to say my app was failing in Jaguar and you pointed out this 
was normal. I figured out a way to test my standalone app on my Panther 
machine (temporarily move the stuff in /Library/Frameworks) and it works 
fine (except for another problem I'm about to post about). So I simply 
documented the Jaguar problem and went on.
...

> > P.S. I finally got my extra files properly installed by adding to the
> > bundlebuilder script such that it:
> > - unpacks Modules.zip
> > - copies the two packages in question (overwriting the incomplete
> > versions that were already installed; I plan to try excluding them and
> > leaving Modules.zip alone).
> > The results work on my machine but break on a Jaguar machine is noted
> > above with True undefined -- as if it's running the python on that
> > machine instead of the framework python that's in the application
> > package.
> 
> If you exclude a module/package (so it won't go in Modules.zip), and 
> then include it (either as a resource or a module, it doesn't really 
> matter) then this problem should be solved.  No need to modify the code 
> for bundlebuilder.

Yes, I tried this and found that it greatly increased the # of modules 
that had to be explicitly listed for inclusion. The only way I found of 
dealing with that was iteratively including the modules I got errors 
for, running the program and repeating as new module errors were 
reported. After a few rounds of this I gave up and went back to my old 
system of unpacking Modules.zip.

> That said, I have personally modified a copy of bundlebuilder that 
> fixes the aforementioned symlink bug, and adds a slew of additional 
> features with regard to embedding modules and frameworks (it actually 
> scans your extensions and automatically locates any dylibs/frameworks 
> they depend on.. rewrites all of their headers and chunks them into the 
> app bundle).  I'll clean this up and release it someday if people bug 
> me about it on a semi-regular basis :)

That sounds wonderful.

Thanks for all your help!

-- Russell