On Thu, Dec 10, 2009 at 9:41 AM, Daniel Hoolihan <d...@archoo.net> wrote:
> i completely agree with "lets keep options to a minimum" and i
> appreciate the comment "we shouldn't mind code complexity"
> so .. might i suggest that -m be overloaded?? so both of these would work..
>
Good morning!
i've changed it to -ZZZ and --read-message-from-file.
Just kidding.
The original justification for using "-M" was "-M is an alternate form of
-m," and my instinct says to use -M over -i/--infile (which i find a bit
confusing in this context because the committed files are the "input
files"). Jeremy's suggestion of --message-file is a good alternative to
--comment-file, and seems more intuitive to me.
> fossil commit -m ./file.txt would
> fossil commit -m "commit message"
> essentially -m <value> would first check to see if <value> resolves to a
> valid file, otherwise take <value> to be the comment text..
>
This would actually more than double the code needed, i think. Currently i
use blob.c::blob_read_from_file() to populate the comment string from, and
that function calls fossil_panic() (i.e. exit app) if the file doesn't
exist:
if( zComment ){ /* -m TEXT */
blob_zero(&comment);
blob_append(&comment, zComment, -1);
}else if( zCommentFile ){ /* -M FILENAME */
blob_zero(&comment);
blob_read_from_file(&comment, zCommentFile);
}else{ /* use editor */
prepare_commit_comment(&comment);
}
We'd need to merge the first 'if' and 'else if', and then add the logic to
test for file existence (or just try to open it), and fall back to using
zComment as a -m string if we can't find/open the file. While that would
work 99.9% of the time, it would fail with unexpected results when a commit
message string really does resolve to a readable file. i honestly don't
think it would ever happen, but the possibility is there:
f commit -m /etc/hosts foo.c
Not tragic, but mildly disturbing to look at in the commit log (and may
require a purge of the artifact, depending on the security/sensitivity
level of the system it was pulled from). On a similar note:
f commit -m /etc/sudoers foo.c
would fail (via fossil_panic()) for non-root users because /etc/sudoers is
mode 0440 on most systems.
Again, far-fetched but possible.
i'm with Jeremy on this one: -M/--message-file. That said, Richard's been
interestingly quiet throughout this conversation, which leads me to suspect
that he's hacking away at some clever alternative which will make all this
moot :).
--
----- stephan beal
http://wanderinghorse.net/home/stephan/
_______________________________________________
fossil-users mailing list
fossil-users@lists.fossil-scm.org
http://lists.fossil-scm.org:8080/cgi-bin/mailman/listinfo/fossil-users
