Bug 636856

Created attachment 449222 [details]
File: backtrace

Hi Seth!

I can reproduce this crash regularly by, in bash:

export PYTHONINSPECT=YES
yum search *spe*|less

I know, I know, "to avoid this problem, stop doing that!!!"  ;)


Now the actual bug may lie outside of yum, and the crash is avoided by 
export PYTHONINSPECT=

but I thought I'd let you know in case it helps you or others tracking down brittleness in Fedora. Or maybe you have advice...

Is there a better way to have python scripts conclude by dropping down into the python interpreter? Documentation suggests the use of script headers like

#!/usr/bin/env python -i

but in Fedora such fail with "python -i" not found. Similar to NetBSD discussion
http://mail-index.netbsd.org/netbsd-users/2008/11/09/msg002384.html

Yum is good. Thanks for yum! Best wishes to you and yours!

For reference on PYTHONINSPECT, see http://docs.python.org/dev/using/cmdline.html#envvar-PYTHONINSPECT

If a user has set PYTHONINSPECT in their environment (or used "-i") it suggests that he/she is a technically-proficient user who wants to debug a failing script.

There are some scripts of the form:
  try:
      (main logic of the script)
  except:
      # Hide all exceptions from the user
      sys.exit(1)
for which detection of PYTHONINSPECT might be a good idea.

But I don't think that's relevant to the yum case, which is more of:

    run main logic, get exit_code
    sys.exit(exit_code)

There may be a case for stripping PYTHONINSPECT from the environment, to handle the case where it got accidentally inherited.

(I hope I'm interpreting this bug correctly)

Perhaps I should have said:

"Hi, I'm VERY new to python. I notice that when I turn the environmental variable PYTHONINSPECT on in my global environment by putting in .bashrc,
export PYTHONINSPECT=Yes

a number of well-known packaged fedora applications, when they end normally, cause abrt to complain.  Now when my simple executable scripts end, they don't throw errors. But I don't yet use nor understand python error trapping -- I'm mostly doing mathematical calculations, not managing human input or hardware states. So ...

What I don't know is -- because I don't yet understand python error trapping, nor how abrt does what it does -- whether:

(1) I'm uncovering a programming error in the application which should be reported.  
    <-- I know this may not be true.

(2) I'm uncovering that abrt is too sensitive and complaining when it shouldn't.
    <-- This might be the best way to do this.

(3) I should just set PYTHONINSPECT just when I need it. (And maybe this should go into the python documentation.)
            <-- I'm happy with this answer (3) by the way.

The error lies with ABRT and not yum; 
it's my understanding that a fix is in git.
See Bug 636913.

*** This bug has been marked as a duplicate of bug 636913 ***