On Tue, 2011-12-13 at 16:40 +0100, Hartmut Goebel wrote:
> Am 11.12.2011 14:42, schrieb Giovanni Bajo:
> >> - binchache is going to $HOME/.... (already implemented, but need to
> >> be changed)
> > OK but please fix it as previously discussed (I don't recall al the
> > details now). I'm specifically worried about this.
> 
> This is documented in http://www.pyinstaller.org/ticket/307. We may need 
> to finish this discussion prior to implementing.

What we agreed upon in that ticket is still OK. For Linux, I'm OK with
both using the home directory or /var/cache, but the latter is a better
place for distributions so that they can have a cleanup script to run
when the package is uninstalled.

> >> - find import in extension modules (this would allow some hooks to be
> >> removed)
> > What's your plan here?
> 
> It's just an idea yet, Martin had (http://www.pyinstaller.org/ticket/438).

I don't think there's a good way to do it. The closest thing that comes
to my mind is issuing a warning if the module has a DLL import for a
PyImport* function, but I don't know how effective such a warning would
be.

> >> - infos about how to package as an rpm (ReadMe-Pacakgingg.txt)
> > This is something that it is better left to the distros, since each
> > distro will have different requirements. Again, we must provide a
> > reasonable application (read-only source directory, UNIX-friendly
> > paths,e tc.), but the rest is up to them.
> Same reason as above: Increase user base. A Readme eases the work for 
> the packagers since they get informations where are the places (or at 
> least some of them) they need to look at. E.g. patching pathes in 
> `PyInstaller/__init__.py` It's a service from us developers :-)

Ah sorry, I somehow missed that you were simply suggesting
documentation. That's fine by me.

> >> - idea, internal: use Python zlib instead of own copy
> > We can't because we need to access zlib before python itself is loaded.
> > I would evaluate using a more modern compression library instead (eg:
> > xz) with higher compression ratios, assuming decompression speed is not
> > impacted too much.
> 
> Sorry, I meant libz.so, not the python module zlib.

That would be a Linux only change, then. It might be OK, I think it's
safe to assum that libz.so will always be available and will have a
stable API.
-- 
Giovanni Bajo   ::  [email protected]
Develer S.r.l.  ::  http://www.develer.com

My Blog: http://giovanni.bajo.it

Attachment: smime.p7s
Description: S/MIME cryptographic signature

Reply via email to