aboutsummaryrefslogtreecommitdiffstats
path: root/.be/bugs/496edad5-1484-413a-bc68-4b01274a65eb/comments/8d927822-eff9-42c4-9541-8b784b3f7db2/body
diff options
context:
space:
mode:
authorW. Trevor King <wking@drexel.edu>2008-11-22 16:15:16 -0500
committerW. Trevor King <wking@drexel.edu>2008-11-22 16:15:16 -0500
commit6d4785e75e1552b3f04b1499fede6fdef2732c39 (patch)
tree7fda56847fe1d7515cac381b6ed414b4afe5a1c9 /.be/bugs/496edad5-1484-413a-bc68-4b01274a65eb/comments/8d927822-eff9-42c4-9541-8b784b3f7db2/body
parent93801606303d79cfe9c1483cd6627cb1b93dedc7 (diff)
downloadbugseverywhere-6d4785e75e1552b3f04b1499fede6fdef2732c39.tar.gz
Created and fixed bug 496edad5-1484-413a-bc68-4b01274a65eb.
I figured out why Arch was complaining. For non-Arch users, file system access has been tweaked a bit see the BugDir doc string for details. Also, you should now set BugDir.rcs instead of .rcs_name. .rcs_name automatically tracks changes in .rcs (the reverse of the previous situation), so read from whichever you like.
Diffstat (limited to '.be/bugs/496edad5-1484-413a-bc68-4b01274a65eb/comments/8d927822-eff9-42c4-9541-8b784b3f7db2/body')
-rw-r--r--.be/bugs/496edad5-1484-413a-bc68-4b01274a65eb/comments/8d927822-eff9-42c4-9541-8b784b3f7db2/body29
1 files changed, 29 insertions, 0 deletions
diff --git a/.be/bugs/496edad5-1484-413a-bc68-4b01274a65eb/comments/8d927822-eff9-42c4-9541-8b784b3f7db2/body b/.be/bugs/496edad5-1484-413a-bc68-4b01274a65eb/comments/8d927822-eff9-42c4-9541-8b784b3f7db2/body
new file mode 100644
index 0000000..dfcf82c
--- /dev/null
+++ b/.be/bugs/496edad5-1484-413a-bc68-4b01274a65eb/comments/8d927822-eff9-42c4-9541-8b784b3f7db2/body
@@ -0,0 +1,29 @@
+I was having problems with `python test.py bugdir` with the Arch
+backend. Commits were failing with `archive not registered'.
+
+Adding some trace information to arch.Arch._rcs_init() and
+._rcs_cleanup() (the traceback module is great :p), I found
+that the problem was coming from bugdir.BugDir.guess_rcs().
+
+The Arch backend deletes any auto-created archives when it is cleaned
+up (RCS.__del__ -> RCS.cleanup -> Arch._rcs_cleanup). This means that
+whatever instance is used to init the archive in guess_rcs() must be
+kept around. I had been doing:
+ * installed_rcs() -> Arch-instance-A
+ * Arch-instance-A.init()
+ * store Arch-instnance-A.name as bugdir.rcs_name
+ * future calls to bugdir.rcs get new instance Arch-instance-B
+ * eventually Arch-instance-A cleaned up
+ * archive dissapears & tests crash
+
+I switched things around so .rcs is the `master attribute' and
+.rcs_name follows it. Now just save whichever rcs you used to init
+your archive as .rcs.
+
+In order to implement the fix, I had to tweak the memory/file-system
+interaction a bit. Instead of saving the settings *every*time* a
+setting_property changed, we now save only if the .be file exists.
+This file serves as a 'file-system-bugdir-active' flag. Before it is
+created (e.g., by a .save()), the BugDir lives purely in memory, and
+can freely go about configuring .rcs, .rcs_name, etc until it get's
+to the point where it's ready to go to disk.