Giovanni Bajo <[EMAIL PROTECTED]> writes:
> Daniel Diniz wrote:
> > Just FWIW, this fix solved a rather nasty build crash for me.
> Wow, very well then!
> Since I'm curious, do you have any idea on what kind of code produce that
> STORE_DEREF in the context of an import?
Hi, Giovanni
It seems that in my case, the problem was this line:
import wx.lib.plot
That does the (crash) trick for me (Python 2.4.2 [MSC v.1310 32 bit (Intel)] on
win32 Win98). I'd be able to figure this sooner if I didn't expect STORE_DEREFs
to be generated all the time, and they are. I used the following (mf.py,
scan_code()):
elif op in STORE_OPS:
++ if op == STORE_DEREF:
++ print "STORE_DEREF"
++ print op, co.co_names[oparg], conditional, curline
pass
But not all STORE_DEREFs cause the problem (and I was flagging all occurrences,
so... :)). Using the following code
++ print mod.__doc__, mod.__name__
mod.xref(importer)
for name, isdelayed, isconditional in mod.imports:
in mf.py's analyze_r, the last line before a crash is "None ctypes". And
"ctypes" only appears in the log when I import wx.lib.plot. With the OP's patch
and the reporting code above, the log goes on and the build finishes, leaving
the next lines as possible clues:
None ctypes
STORE_DEREF
137 sys 0 0
STORE_DEREF
I hope this helps somehow :)
Best regards,
Daniel
_______________________________________________
PyInstaller mailing list
[email protected]
http://lists.hpcf.upr.edu/mailman/listinfo/pyinstaller
