Skip to contentSkip to conversation

check-ignore: Add option to ignore index contents

PATCH In review Dave Williams 2 participants 4 messages Jun 15, 2016 Documentation

lore
Share & tools
lore
#0 lore raw

PATCH v3 1/1 check-ignore: Add option to ignore index contents

check-ignore currently shows how .gitignore rules would treat untracked paths. Tracked paths do not generate useful output. This prevents debugging of why a path became tracked unexpectedly unless that path is first removed from the index with git rm --cached <path>.

This option (-i, --no-index) simply by-passes the check for the path being in the index and hence allows tracked paths to be checked too.

Whilst this behaviour deviates from the characteristics of git add and git status its use case is unlikely to cause any user confusion.

Test scripts are augmented to check this option against the standard ignores to ensure correct behaviour.

---
In V3 I have taken on board comments from Junio Hamano and Eric Sunshine from V2 ($gmane/233660).

In particlar Junio queried my approach in builtin/git-check-ignores.c that bypassed the functions that check a path is in the index as well as avoiding reading the index in the first place.

In my view removing the bypass makes the code slightly less clear, relies on the_index being initialized and the functions using it to exit quickly when it has no content. Nevertheless I have bowed to his better domain knowledge and after undertaking brief analysis to check the assumptions I have removed the extra conditional steps. This has simplified the patch. The revised code appears to have the same behaviour as before and passes the test script (t/t00009-ignors.sh).

Regarding the test script I have tidied up the variables containing the separate option switches so they dont contain leading spaces, instead I have added spaces as needed when formatting the command line. This not only improves my patch but also the existing code which was a little inconsistent in this respect.

Finally I have rebased from the latest commmit on master to pick up unrelated recent changes made to builtin/check-ignores.c and updated my code to be consistent with this.

Hopefully I have put these patch notes in the right place now! Let me know if not.

Dave

 Documentation/git-check-ignore.txt |  7 ++++
 builtin/check-ignore.c             |  6 ++-
 t/t0008-ignores.sh                 | 75 +++++++++++++++++++++++++++++++++-----
 3 files changed, 77 insertions(+), 11 deletions(-)
Documentation/git-check-ignore.txt +7 -0
index d2df487..96c591f 100644
@@ -45,6 +45,13 @@ OPTIONS
 	not be possible to distinguish between paths which match a
 	pattern and those which don't.
 
+-i, --no-index::
+	Don't look in the index when undertaking the checks. This can
+	be used to debug why a path became tracked by e.g. `git add .`
+	and was not ignored by the rules as expected by the user or when
+	developing patterns including negation to match a path previously
+	added with `git add -f`.
+
 OUTPUT
 ------
 
builtin/check-ignore.c +4 -2
index 25aa2a5..f58f384 100644
@@ -5,7 +5,7 @@
 #include "pathspec.h"
 #include "parse-options.h"
 
-static int quiet, verbose, stdin_paths, show_non_matching;
+static int quiet, verbose, stdin_paths, show_non_matching, no_index;
 static const char * const check_ignore_usage[] = {
 "git check-ignore [options] pathname...",
 "git check-ignore [options] --stdin < <list-of-paths>",
@@ -24,6 +24,8 @@ static const struct option check_ignore_options[] = {
 		 N_("terminate input and output records by a NUL character")),
 	OPT_BOOL('n', "non-matching", &show_non_matching,
 		 N_("show non-matching input paths")),
+	OPT_BOOL('i', "no-index", &no_index,
+		 N_("ignore index when checking")),
 	OPT_END()
 };
 
@@ -157,7 +159,7 @@ int cmd_check_ignore(int argc, const char **argv, const char *prefix)
 		die(_("--non-matching is only valid with --verbose"));
 
 	/* read_cache() is only necessary so we can watch out for submodules. */
-	if (read_cache() < 0)
+	if (!no_index && read_cache() < 0)
 		die(_("index file corrupt"));
 
 	memset(&dir, 0, sizeof(dir));
t/t0008-ignores.sh +66 -9

163 lines not on this page · raw message

Signature · 1 line
1.8.4.rc3
#1 lore raw
Quoted 7 lines · Dave Williams

check-ignore currently shows how .gitignore rules would treat untracked paths. Tracked paths do not generate useful output. This prevents debugging of why a path became tracked unexpectedly unless that path is first removed from the index with git rm --cached <path>.

This option (-i, --no-index) simply by-passes the check for the path being in the index and hence allows tracked paths to be checked too.

Now the long option name is "--no-index", it makes me wonder if "-i" is a good synonym for it, and the longer I stare at it, the more certain I become convinced that it is a bad choice.

Do we even need a short-and-sweet one-letter option for this? I'd prefer starting with only the long option.

I came up with a squashable tweak to remove "-i" on top of this patch; will tentatively queue this patch with it to 'pu'.

Quoted 4 lines

In particlar Junio queried my approach in builtin/git-check-ignores.c that bypassed the functions that check a path is in the index as well as avoiding reading the index in the first place.

Thanks.

I think this version is cleaner without those "if (!no_index)" used for special casing in the codeflow. An empty index is a valid state, in which the in-core index starts when any git program begins. Not reading the index should be the only thing necessary to mimick the state in which nothing has been added to the index, and if that is not the case, we have found a bug in the existing code.

Quoted 5 lines

Regarding the test script I have tidied up the variables containing the separate option switches so they dont contain leading spaces, instead I have added spaces as needed when formatting the command line. This not only improves my patch but also the existing code which was a little inconsistent in this respect.

Yeah, that's very much appreciated.

Quoted 8 lines

Finally I have rebased from the latest commmit on master to pick up unrelated recent changes made to builtin/check-ignores.c and updated my code to be consistent with this.

Hopefully I have put these patch notes in the right place now! Let me know if not.

Dave

Nicely done.

#2 lore raw
Quoted 4 lines · Junio C Hamano

Now the long option name is "--no-index", it makes me wonder if "-i" is a good synonym for it, and the longer I stare at it, the more certain I become convinced that it is a bad choice.

I had originally called it "--ignore-index" at which point "-i" made more sense!

Dave

#3 lore raw
Quoted 7 lines · Dave Williams
Quoted 4 lines · Junio C Hamano

Now the long option name is "--no-index", it makes me wonder if "-i" is a good synonym for it, and the longer I stare at it, the more certain I become convinced that it is a bad choice.

I had originally called it "--ignore-index" at which point "-i" made more sense!

I know, but with --no-index it no longer does.