Skip to content

Conversation

@vokimon
Copy link

SymbolicReference leaked file descriptor resources when iterating on refs. unittest module complained:

symbolic.py:153: ResourceWarning: unclosed file <_io.TextIOWrapper name='...../gitrepo/.git/packed-refs' mode='rt' encoding='UTF-8'>
symbolic.py:148: ResourceWarning: unclosed file <_io.TextIOWrapper name='....../gitrepo/.git/packed-refs' mode='rt' encoding='UTF-8'>

Byron added a commit that referenced this pull request Jul 27, 2015
@ByronByron merged commit 7c96f58 into gitpython-developers:masterJul 27, 2015
@Byron
Copy link
Member

Thank you very much !

I am afraid there are many more issues of that kind buried in the code as I was wrongly assuming (back in the days) that an object destructor is deterministically called (such as in C++). This led to me failing to explicitly close files.

@ByronByron added this to the v1.0.2 - Fixes milestone Jul 27, 2015
@vokimon
Copy link
Author

what about starting with a grep for 'open'?

@Byron
Copy link
Member

One will have to spread these efforts among GitDB and SMMap as well. Both should have mechanisms to release resources, but I doubt this works properly.

For now the best way to workaround these resource related issues is to use a GitCmdObjectDB, at least when working with non-ref/object data.

EliahKagan added a commit to EliahKagan/GitPython that referenced this pull request Nov 3, 2023
This removes a comment noting that a try-finally block had been present (or had been intended), but was removed because some version of Python had imposed a limitation on yield appearing in try-finally. That comment was obsolete as of 58c5b99 (gitpython-developers#326), which wrapped the relevant code in a with-statement, because: 1. Since then, the cleanup is done in a manner equivaent to try-finally. 2. It turned out, as noted in that PR, that cleanup had not always been done automatically. (This was contrary to the prediction given in the comment.) 3. At some point before that, the limitation that had prevented the use of try-finally no longer affected any supported version of Python. Specifically, it appears the only limitation that this could be was the limitation lifted in Python 2.5, where along with the addition of the close() method which causes try-finally to be called (and is itself called when a generator object is finalized), yield in a try-block with an associated finally-block became permitted, since the call to close() was sufficient to run the finally-block (by raising GeneratorExit). For details, see: https://docs.python.org/3/whatsnew/2.5.html#pep-342-new-generator-features (This obsolete comment was one of the things I discovered while working on gitpython-developers#1725, but I didn't include this change there, having not yet looked into the history of the code enough to be sure.)
EliahKagan added a commit to EliahKagan/GitPython that referenced this pull request Nov 3, 2023
This removes a comment noting that a try-finally block had been present (or had been intended), but was removed because some version of Python had imposed a limitation on yield appearing in try-finally. That comment was obsolete as of 58c5b99 (gitpython-developers#326), which wrapped the relevant code in a with-statement, because: 1. Since then, the cleanup is done in a manner equivaent to try-finally. 2. It turned out, as noted in that PR, that cleanup had not always been done automatically. (This was contrary to the prediction given in the comment.) 3. At some point before that, the limitation that had prevented the use of try-finally no longer affected any supported version of Python. Specifically, it appears the only limitation that this could be was the limitation lifted in Python 2.5, where along with the introduction of close(), which is automatically called when a generator object is finalized, it became permitted for yield to appear in a try-block with an associated finally-block, on the grounds that calling close() runs the finally-block (by raising GeneratorExit). For details, see: https://docs.python.org/3/whatsnew/2.5.html#pep-342-new-generator-features (This obsolete comment was one of the things I discovered while working on gitpython-developers#1725, but I didn't include this change there, having not yet looked into the history of the code enough to be sure.)
Sign up for freeto join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Development

Successfully merging this pull request may close these issues.

2 participants

@vokimon@Byron