Tor Software Error - The Tor software encountered an internal bug. Please report the following error message: "microdesc_cache_rebuild(): Bug: Discontinuity in position in microdescriptor cache.By my count, I'm at 1760119, but I should be at 1760185
"
Attached you see the log of Vidalia (what is left).
It's a bridge (0.2.4.12) running on Windows 7.
I never saw this error before.
My normal reaction is restart the damn thing, but the bug says I should report it. (I don't see errors at all).
Tor has permission to write to the folder it attempts to. I can't see anything locking the file. Tor has created the file previously and there's only one Tor instance running.
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information
Child items
...
Show closed items
Linked items
0
Link issues together to show that they're related.
Learn more.
I let Tor repeat this error over and over again and did not stop it as is wasn't showing any connectivity/usage problems.
Today I launched it again and Vidalia logged the errors again, so I stopped Tor.
I went to %APPDATA%\Tor and removed the following files
cached-microdescs (The one Tor can't write to)
cached-microdescs.new (I saw this file before, removed to have it fresh)
cached-microdescs.tmp (I can't remember to have since such a file before, removed to have it fresh)
History
cached-microdescs created 2012-05-19, changed 2013-04-28, accessed 2013-04-28
cached-microdescs.new created 2012-05-19, changed 2013-05-05, accessed 2013-04-28
cached-microdescs.tmp created 2013-05-04, changed 2013-05-05, accessed 2013-05-04
(So I was correct .tmp was never there before)
All files shared the same access-privileges.
I restarted Tor and it created the first two files. No error since then.
The files created share the same access-privileges and are the same as there were before. Since I had not seen this error during the alpha-phase of branch 0.2.4.x I change the version to 0.2.4.12-alpha.
The strange about this issue is that I upgraded from .11 to .12 on 2013-04-30 and the error appeared on 2013-05-04, never before. Might be that it takes Tor some time to gather enough microdescriptors to exhibit this.
I change to "needs more information" and wait to see if it occurs again in which case I report back or ask if we can close this ticket otherwise.
Trac: Summary: Microdescriptor cache fails to Microdescriptor cache fails (Tor can't write to it) Status: new to needs_information Version: N/Ato Tor: 0.2.4.12-alpha
I wonder, could this be another instance of #2077 (moved) ? If so, does any of the discussion there seem relevant?
Looks like this would be the same issue. I'm not sure what's relevant. However based on the ticket I can give the following information:
File flags for %APPDATA%\Tor (I don't know how they are named in English versions of Windows)
write-protected/read-only: not set
hidden: not set
archiv: set (file can be backed-up)
index content: not set
compress content: not set
encrypt content: not set (grayed out, not available)
I (disabled publishing my descriptor and) started Tor.
[...]
Mai 09 08:23:49.039 [Warning] Error replacing "C:/Users/[username]/AppData/Roaming/tor\cached-microdescs": Permission denied
Mai 09 08:23:49.039 [Warning] Error rebuilding microdescriptor cache: Permission denied
I launched ProcessMonitor (before I launched Tor) to provide information what accesses these files. I'll attach an excerpt of what was logged. Tor could write to .tmp, but could not remove the "cached-microdescs" file. CANNOT DELETE. There's no race between two processes accessing the file(s).
I never had this error prior to 0.2.4.12-alpha and wonder if there was change.
#8037 (moved) is mentioned in the change-log. I don't see what would cause this.
I never saw:
[Warning] Gzip returned an error: stream error
[Warning] write_to_buf failed. Closing connection (fd -1).
in my logs. Those lines are not in the logs I attached (unless I overred them). I can't remember to have seen them.
This is not, as I'd thought, a dup of #2077 (moved) or anything else. It's a duplicate of an old bug that never got a number which we fixed in bdff7e3299d78cd2f7aa4a31e5f57c1aeef5ffa1 but reintroduced in 6905c1f6.