blob: fce51ac322a4f4879d48aa866e3fe9385d051008 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
|
[[!comment format=mdwn
username="http://joeyh.name/"
ip="4.152.246.110"
subject="comment 13"
date="2013-07-27T20:46:01Z"
content="""
I've verified that the (old) code correctly returned no path when it's too long:
<pre>
*Annex.Ssh System.Directory System.Posix.Directory Annex Git.Construct Git.Config System.Posix.Env Utility.Env> eval (newState r') $ changeGitConfig (\c -> c { annexCrippledFileSystem = True }) >> sshInfo (\"fozz@git-annex-markdown.lang.speechmarks.com-fozz_phone.2Dannex.IdWwlXHtSsjVUMcq\", Nothing)
(Nothing,[])
</pre>
In case the problem was somehow related to calculating the length on Android (ie, Data.Bits.Utils broken), I built a small program that calculated the length of the string the same way
[[!format haskell \"\"\"
import Utility.FileSystemEncoding
main = do
let s = \"/data/data/ga.androidterm/tmp/fozz@git-annex-markdown.lang.speechmarks.com-fozz_phone.2Dannex.IdWwlXHtSsjVUMcq\"
let ws = decodeW8 s
print (length ws)
\"\"\"]]
Outputs 110 on Android as expected.
Finally, using git-annex 4.20130721-g002de3e on Android (a newer version than the bug reporter's but with no changes to Annex.Ssh), I manually set up a repository using the same hostname:
<pre>
[remote \"foo\"]
url = ssh://fozz@git-annex-markdown.lang.speechmarks.com-fozz_phone.2Dannex.IdWwlXHtSsjVUMcq
fetch = +refs/heads/*:refs/remotes/foo/*
annex-uuid = dummy
</pre>
And still cannot replicate the bug; as expected it does not use the socket since it's too long:
copy foo (checking foo...) [2013-07-27 16:40:42 EDT] call: ssh [\"-T\",\"fozz@git-annex-markdown.lang.speechmarks.com-fozz_phone.2Dannex.IdWwlXHtSsjVUMcq\",\"git-annex-shell 'inannex' '' 'SHA256E-s29--093429efb0d1427753d1f038f5279ec4df66785a1b2429b3fa5e3a01bcb75bd8' --uuid 111\"]
So, I don't understand how this could have happened. Although my recent changes mean it'll use a 62 byte path max on Android now, which certianly should avoid the problem, even if there's some actual bug here that I cannot reproduce.
"""]]
|
