response

1 files changed, 34 insertions, 0 deletions

diff --git a/doc/bugs/Upload_to_box_very_slow/comment_1_602697f6ea6cfd07764b1e9ed040736c._comment b/doc/bugs/Upload_to_box_very_slow/comment_1_602697f6ea6cfd07764b1e9ed040736c._comment
new file mode 100644
index 000000000..a5e9b69c3
--- /dev/null
+++ b/doc/bugs/Upload_to_box_very_slow/comment_1_602697f6ea6cfd07764b1e9ed040736c._comment
@@ -0,0 +1,34 @@
+[[!comment format=mdwn
+ username="joey"
+ subject="""comment 1"""
+ date="2017-10-07T17:03:56Z"
+ content="""
+I suspect that the problem dealt with in
+[[!commit 2ca1d3cc0134134314649c21822bf1352df52e93]]
+is responsible for this behavior. If many redirects are
+followed before git-annex gives up, it would explain why it's taking so long.
+
+In redirecting from a non-existant file to some index.php file that itself
+doesn't exist, and so on forever, box.com is clearly a horrible, broken,
+very bad webdav server. Given that it's so broken, it's not terribly
+surprising that small changes might expose that or another brokenness in
+different ways. Changing the location of a temp file from a tmp
+subdirectory to the parent directory is not the kind of thing that causes
+30 second delays with non-broken servers.
+
+I have not been able to reproduce the problem here though so that is just a
+theory.
+
+I am not inclined to make random changes to try to work around random
+breakage in box.com. That is a game you lose before you start playing.
+
+It may be that deleting line 132 from Remote/WebDAV.hs avoids
+the problem.
+
+        maybe noop (void . mkColRecursive) (locationParent tmp)
+
+Since the parent directory of the temp file is the top of the webdav
+repository, and so already exists, that line is now unnecessary.
+Since you can reproduce the problem and build from source, can you test
+that change?
+"""]]