CVS import process not very resilient to transient errors
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Launchpad CSCVS |
Triaged
|
Low
|
Unassigned | ||
Launchpad itself |
Triaged
|
Low
|
Unassigned |
Bug Description
Observe the logs of the first two attempts to import the present lp:~naesten/ucon64/trunk (there was previously another vcs-import branch there, which is now named lp:~naesten/ucon64/trunk-broken):
http://
http://
In particular, note that both runs ended with the same type of error, but on different files. In fact, it seems that the second attempt died much earlier on, in a file/revision that had been fetched succussfully in the first attempt!
Such hickups are probably unnavoidable with a system like CVS, so it would probably be wise to make the import process tolerate such errors (at small rates, at least) and just retry a bit later; an appropriate retry interval would probably be somewhere between a few seconds and a few minutes.
Alternatively, perhaps the CVS import process could be made a bit more incremental, so that even if such errors would halt the import process, the next import could start approximately where the failing one left off, like bzr-svn can already do with SVN imports?
Changed in launchpad: | |
status: | New → Triaged |
importance: | Undecided → High |
We're not currently working on this area of LP. If someone wants to improve it we'd love patches and be delighted to mentor folk.