Anders Waldenborg [off-list ref] writes:
From what I can understand it tries to match *both* on the second level
AND the value of .key (trailers.c:token_matches_item)
Yuck, I do not know what were we thinking to design the behaviour
like *that*. Or it may be simply buggy.
$ printf '\na: 1\nb: 2\nc: 3\n' | \
git -c 'trailer.A.key=B' interpret-trailers
B: 1
B: 2
c: 3
I can understand the first one (i.e. "trailer.$name.$var" try to
match $name as case insensitively) but not the second one. There is
not an single rule for "b" trailer, and we should be getting the
same behaviour as the third line, i.e. the key not involved in
rewriting is passed as-is.
Junio C Hamano writes:
Anders Waldenborg [off-list ref] writes:
quoted
From what I can understand it tries to match *both* on the second level
AND the value of .key (trailers.c:token_matches_item)
Yuck, I do not know what were we thinking to design the behaviour
like *that*. Or it may be simply buggy.
quoted
$ printf '\na: 1\nb: 2\nc: 3\n' | \
git -c 'trailer.A.key=B' interpret-trailers
B: 1
B: 2
c: 3
I can understand the first one (i.e. "trailer.$name.$var" try to
match $name as case insensitively) but not the second one. There is
not an single rule for "b" trailer, and we should be getting the
same behaviour as the third line, i.e. the key not involved in
rewriting is passed as-is.
I'm not so sure about that. Matching sometimes needs to happen on .key
too. If this configuration is supposed to be useful (and as "shortcuts"
has been mentioned before and is what tests does, I think it should be):
trailer.rb.key=Reviewed-By
trailer.rb.ifexists=addifdifferent
then matching must look at key, not name. As the value of .key is what
would have been written previously into the message.
E.g:
$ printf "\nReviewed-By: hi\n" | \
git -c "trailer.rb.key=Reviewed-By" \
-c "trailer.rb.ifexists=addifdifferent" \
interpret-trailers --trailer "rb=hi"
Reviewed-By: hi
The opposite case, matching on name in input message I'm not sure where
it is useful.
$ printf "\nrb: hi\n" | \
git -c "trailer.rb.key=Reviewed-By" \
-c "trailer.rb.ifexists=addifdifferent" \
interpret-trailers --trailer "rb=hi"
Reviewed-By: hi
The way I have understood it ".key" can be used for some different
things:
* Freeing up name to be used as a shortcut alias.
* Defining the canonical capitalization when passing through trailers
* Allowing specifying non default separator for that key.
I wonder if those uses could be split up.
So instead of this configuration:
[trailer "rb"]
key="Reviewed-By> "
that does all three of those. The configuration would be:
[trailer "Reviewed-By"]
separator="> "
canonicalize=true
alias=rb
that way the "name" part of the config always is the correct spelling
and capitalization. "alias" could easily be a list of multiple alias if
that is wanted. "alias" could be split into "inputalias" (matched
against when reading trailers from stdin or a commit/tag message) and
"optalias" (matched against when reading --trailer cmdline option, and
%(trailers:key))