[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
cvs bug in "update"
cvs bug in "update"
Tue, 23 Jul 2002 17:45:06 -0500
When doing a cvs update and encountering a "move away" warning with one file
and a "checksum failure after patch" on another file, the unpatchable files
cannot be refetched in the second pass. The combination of these two warnings
produces undesirable results in that files that can normally be updated get
To reproduce this, both warnings must be created. You must use a
case-insensitive filesystem (i.e., Windows)
1) make a change elsewhere that produces the "checksum failure after patch".
I don't know how I did this, but it appears that the files showing that
error are in DOS mode on my computer and unix mode in the repository.
2) change the case of a file such that you would get the "move away" problem.
You can achieve this easily by changing the case in the Entries file.
Reading through some code, it makes sense to not retrieve files that encounter
the "move away" warning. However, it also says that repeating the update
takes care of the problem. That's wrong, but that's an topic for another
discussion (i.e., this behavior may actually be desired).
I believe that if failure_exit is initialized to 0 at the beginning of
get_server_responses(), the problem is fixed. Because update has a 2-pass
scenario when patching fails, the second pass will see the failure_exit
condition from the first pass and fail out of start_server() right after the
connection is made.
Unfortunately, I haven't tested very thoroughly. I would really appreciate
some insight on this. Are there other places where start_server is called
multiple times? Perhaps the second connect isn't really needed and is just
there due to laziness. I'm eager to help in any way that I can.
, , , | Duane Bronson
/|/|/| , | address@hidden
( ( ( |/| | http://www.nerdlogic.com/
\ ( | | 453 Washington St. #4A, Boston, MA 02111
| / | (617) 515-2909
|[Prev in Thread]
||[Next in Thread]|
- cvs bug in "update",
Duane Bronson <=