Charles-François Natali neolo...@free.fr added the comment:
I still don't like the idea of adding such a kludge to work around OS
bugs, so I'd suggest closing as won't fix.
--
___
Python tracker rep...@bugs.python.org
Antoine Pitrou pit...@free.fr added the comment:
+1 for closing.
--
nosy: +pitrou
___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue14702
___
___
Andrew McNabb amcn...@mcnabbs.org added the comment:
Which one-line fix do you propose?
Doing a stat(/net/prodigy/tmp) before mkdir(/net/prodigy/tmp) is an
extremely simple workaround.
Of course, I would love to see clear documentation of how the kernel is defined
to behave in this
Martin v. Löwis mar...@v.loewis.de added the comment:
Of course, I would love to see clear documentation of how the kernel
is defined to behave in this situation. It would certainly be
intuitive for mkdir(/net/prodigy/tmp) to force a mount of
/net/prodigy, but defined behavior isn't
Andrew McNabb amcn...@mcnabbs.org added the comment:
Maybe I'm confused, but the presence of /net/prodigy is *not*
the issue here, and what gets mounted is *not* /net/prodigy,
but /net/prodigy/tmp (do mount to confirm or dispute).
No, /net/prodigy is the mountpoint in this case:
Charles-François Natali neolo...@free.fr added the comment:
I see no evidence that this is a bug in Linux,
stat(/net/prodigy, {st_mode=S_IFDIR|0755, st_size=4096, ...}) = 0
mkdir(/net/prodigy/tmp, 0777) = -1 EACCES (Permission denied)
As you can see, a stat() is already done on
Hynek Schlawack h...@ox.cx added the comment:
stat(/net/prodigy, {st_mode=S_IFDIR|0755, st_size=4096, ...}) = 0
mkdir(/net/prodigy/tmp, 0777) = -1 EACCES (Permission denied)
As you can see, a stat() is already done on /net/prodigy.
To be fair, that shouldn’t trigger a mount.
Martin v. Löwis mar...@v.loewis.de added the comment:
stat(/net/prodigy, {st_mode=S_IFDIR|0755, st_size=4096, ...}) = 0
mkdir(/net/prodigy/tmp, 0777) = -1 EACCES (Permission denied)
As you can see, a stat() is already done on /net/prodigy.
To be fair, that shouldn’t trigger a
Hynek Schlawack h...@ox.cx added the comment:
stat(/net/prodigy, {st_mode=S_IFDIR|0755, st_size=4096, ...}) = 0
mkdir(/net/prodigy/tmp, 0777) = -1 EACCES (Permission denied)
As you can see, a stat() is already done on /net/prodigy.
To be fair, that shouldn’t trigger a mount.
Charles-François Natali neolo...@free.fr added the comment:
Alright, closing for good then.
Andrew, if you want to get this fixed, you should report this to the autofs
folks, because it's definitely not a Python bug.
--
stage: - committed/rejected
status: open - closed
Andrew McNabb amcn...@mcnabbs.org added the comment:
I see no evidence that this is a bug in Linux, and I think it's ridiculous to
close it when a trivial one-line fix is available. I won't reopen it because
it's obvious no one wants to address this. :(
--
Andrew McNabb amcn...@mcnabbs.org added the comment:
I posted a bug report with the kernel here:
https://bugzilla.kernel.org/show_bug.cgi?id=43262
--
___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue14702
Charles-François Natali neolo...@free.fr added the comment:
I guess this is the magic in mkdir -p:
mkdir(expert, 0755) = -1 EACCES (Permission denied)
chdir(expert) = 0
mkdir(tmp, 0755) = -1 EEXIST (File exists)
I'm not sure
Hynek Schlawack h...@ox.cx added the comment:
I agree, closing.
Thank you for your valuable assistance Andrew!
--
resolution: - wont fix
stage: needs patch -
status: open - closed
___
Python tracker rep...@bugs.python.org
Andrew McNabb amcn...@mcnabbs.org added the comment:
This isn't fixed. All of the examples I've given were with a 3.3.0 kernel.
Doing a stat would be a fix.
--
status: closed - open
___
Python tracker rep...@bugs.python.org
Hynek Schlawack h...@ox.cx added the comment:
If I read it correctly, we _do_ a stat:
stat(/net/prodigy, {st_mode=S_IFDIR|0755, st_size=4096, ...}) = 0
mkdir(/net/prodigy/tmp, 0777) = -1 EACCES (Permission denied)
ISTM that this is a different problem altogether, or am I missing
Hynek Schlawack h...@ox.cx added the comment:
My suspicion re the race condition is that a chdir waits for autofs to mount
but mkdir doesn't. You could check that yourself if a chdir/sleep would help.
--
___
Python tracker rep...@bugs.python.org
Andrew McNabb amcn...@mcnabbs.org added the comment:
Hmm. Maybe there's a difference between doing stat('/net/prodigy') vs.
stat('/net/prodigy/tmp'). Just a guess, but maybe the former can succeed before
the mount completes, but the latter has to wait for the mount to complete.
--
Andrew McNabb amcn...@mcnabbs.org added the comment:
By the way, if my hunch about the difference in stat of '/net/prodigy' vs.
'/net/prodigy/tmp' is correct, then this would explain why makedirs on deeper
directories work. Specifically, one of the shallower stat calls would force the
mount
Charles-François Natali neolo...@free.fr added the comment:
Yes, creating the directories in a bottom-up way (i.e. '/net', '/net/prodigy',
'/net/prodigy/foo') could maybe avoid this problem.
But this is definietely an autofs bug, and there are probably many other places
where such code might
Hynek Schlawack h...@ox.cx added the comment:
I sense a very easy fix: just do a stat() on the final path before we do
anything (and quickly return if present). Moving current makedirs into a
private _makedirs that gets called if the stat fails.
I'm still feeling uneasy about that though. It
Martin v. Löwis mar...@v.loewis.de added the comment:
I agree that the proposed solutions are hacks, and still propose to close this
as won't fix.
I think this is a bug in Linux. mkdir(/net/prodigy/tmp) ought to trigger the
mount, then fail with EEXIST.
--
nosy: +loewis
Hynek Schlawack h...@ox.cx added the comment:
Andrew, are you still with us?
--
___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue14702
___
___
Andrew McNabb amcn...@mcnabbs.org added the comment:
Andrew, are you still with us?
I'm here, but it's been a busy few weeks. I'll see if I can spend some time on
this today.
--
___
Python tracker rep...@bugs.python.org
Andrew McNabb amcn...@mcnabbs.org added the comment:
Some interesting information. If I do `os.mkdir('/net/prodigy/tmp')`, it gives
OSError: [Errno 13] Permission denied: '/net/prodigy/tmp'. However, if I
instead do `os.mkdir('/net/prodigy/tmp/hi')`, it succeeds. (Note that I'm being
careful
Changes by Andrew McNabb amcn...@mcnabbs.org:
Added file: http://bugs.python.org/file25611/mkdir-p.out
___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue14702
___
Changes by Andrew McNabb amcn...@mcnabbs.org:
Added file: http://bugs.python.org/file25612/makedirs.out
___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue14702
___
Hynek Schlawack h...@ox.cx added the comment:
I guess this is the magic in mkdir -p:
mkdir(expert, 0755) = -1 EACCES (Permission denied)
chdir(expert) = 0
mkdir(tmp, 0755) = -1 EEXIST (File exists)
I'm not sure how I feel about
Hynek Schlawack h...@ox.cx added the comment:
David: What do you mean? I'd just catch EPERM and look whether the directory
appeared in the meantime. If it exists, continue. If not: re-raise. Am I
missing something? I may have a look at GNU mkdir if it doesn't help.
--
Charles-François Natali neolo...@free.fr added the comment:
To me, this doesn't look like a os.makedirs() bug, but rather an autofs one
(didn't know it was still in use).
I don't like the idea of adding such a kludge (i.e. catching EPERM and checking
whether the directory appeared in between).
Hynek Schlawack h...@ox.cx added the comment:
Charles, I don't think you can blame autofs here. The problem at hand is that
makedirs() never checks whether the directory exists (that would trigger the
mount too I presume). Instead, it tries a mkdir and looks if it gets an EEXIST.
If you try
Charles-François Natali neolo...@free.fr added the comment:
Charles, I don't think you can blame autofs here. The problem at hand is that
makedirs() never checks whether the directory exists (that would trigger the
mount too I presume).
Yes, it does. Have a look at line 148:
if head
R. David Murray rdmur...@bitdance.com added the comment:
Hynek: you said just like EEXIST, which doesn't check to see if the directory
exists before continuing, thus my confusion.
If mkdir -p does a stat first, then changing the makedirs algorithm to match is
probably not a bad idea for OS
Hynek Schlawack h...@ox.cx added the comment:
Charles, I don't think you can blame autofs here. The problem at hand is
that makedirs() never checks whether the directory exists (that would
trigger the mount too I presume).
Yes, it does. Have a look at line 148:
if head and tail
R. David Murray rdmur...@bitdance.com added the comment:
Hynek: it doesn't seem like that would work, since legitimate EPERM errors are
much more likely.
--
nosy: +r.david.murray
___
Python tracker rep...@bugs.python.org
New submission from Andrew McNabb amcn...@mcnabbs.org:
When a os.makedirs is used under an autofs directory, it crashes. For example,
on my machine, `os.makedirs('/net/prodigy/tmp')` crashes with the following
traceback:
Traceback (most recent call last):
...
File
Hynek Schlawack h...@ox.cx added the comment:
As makedirs in 3.x doesn’t handle EPERM and is otherwise the same, I presume
the error is there as well.
I also presume, that after the failed makedirs(), the directory is mounted? I'd
just handle the error just we handle EEXIST in 3.x now.
Changes by Hynek Schlawack h...@ox.cx:
--
keywords: +easy
___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue14702
___
___
Python-bugs-list mailing
38 matches
Mail list logo
