# [PATCH v2] pathspec: honor `PATHSPEC_PREFIX_ORIGIN` with empty prefix 5 messages from 2017-04-04 to 2017-04-12. Participants: Patrick Steinhardt, Brandon Williams, Duy Nguyen, Junio C Hamano. Thread: https://gitlist.dev/t/45603 ## Patrick Steinhardt, 2017-04-04 09:16 Subject: [PATCH v2] pathspec: honor `PATHSPEC_PREFIX_ORIGIN` with empty prefix Message-ID: <4650e1396a5ad006b2b4f355685e3c59efd554d6.1491297393.git.ps@pks.im> URL: https://gitlist.dev/e/4650e1396a5ad006b2b4f355685e3c59efd554d6.1491297393.git.ps%40pks.im ``` Previous to commit 5d8f084a5 (pathspec: simpler logic to prefix original pathspec elements, 2017-01-04), we were always using the computed `match` variable to perform pathspec matching whenever `PATHSPEC_PREFIX_ORIGIN` is set. This is for example useful when passing the parsed pathspecs to other commands, as the computed `match` may contain a pathspec relative to the repository root. The commit changed this logic to only do so when we do have an actual prefix and when literal pathspecs are deactivated. But this change may actually break some commands which expect passed pathspecs to be relative to the repository root. One such case is `git add --patch`, which now fails when using relative paths from a subdirectory. For example if executing "git add -p ../foo.c" in a subdirectory, the `git-add--interactive` command will directly pass "../foo.c" to `git-ls-files`. As ls-files is executed at the repository's root, the command will notice that "../foo.c" is outside the repository and fail. Fix the issue by again using the computed `match` variable when `PATHSPEC_PREFIX_ORIGIN` is set and global literal pathspecs are deactivated. Note that in contrast to previous behavior, we will now always call `prefix_magic` regardless of whether a prefix is actually set. But this is the right thing to do: when the `match` variable has been resolved to the repository's root, it will be set to an empty string. When passing the empty string directly to other commands, it will result in a warning regarding deprecated empty pathspecs. By always adding the prefix magic, we will end up with at least the string ":(prefix:0)" and thus avoid the warning. Signed-off-by: Patrick Steinhardt --- This is the second version of [1]. It fixes a bug catched by Brandon when the pathspec is resolved to the empty string and improves the test a bit to actually catch this issue. [1]: http://public-inbox.org/git/1556910880cfce391bdca2d8f0cbcb8c71371691.1491206540.git.ps@pks.im/T/#u pathspec.c | 2 +- t/t3701-add-interactive.sh | 22 ++++++++++++++++++++++ 2 files changed, 23 insertions(+), 1 deletion(-) diff --git a/pathspec.c b/pathspec.c index 303efda83..3079a817a 100644 --- a/pathspec.c +++ b/pathspec.c @@ -505,7 +505,7 @@ static void init_pathspec_item(struct pathspec_item *item, unsigned flags, * original. Useful for passing to another command. */ if ((flags & PATHSPEC_PREFIX_ORIGIN) && - prefixlen && !get_literal_global()) { + !get_literal_global()) { struct strbuf sb = STRBUF_INIT; /* Preserve the actual prefix length of each pattern */ diff --git a/t/t3701-add-interactive.sh b/t/t3701-add-interactive.sh index f9528fa00..2ecb43a61 100755 --- a/t/t3701-add-interactive.sh +++ b/t/t3701-add-interactive.sh @@ -436,6 +436,28 @@ test_expect_success 'add -p handles globs' ' test_cmp expect actual ' +test_expect_success 'add -p handles relative paths' ' + git reset --hard && + + echo base >relpath.c && + git add "*.c" && + git commit -m relpath && + + echo change >relpath.c && + mkdir -p subdir && + git -C subdir add -p .. 2>error <<-\EOF && + y + EOF + + test_must_be_empty error && + + cat >expect <<-\EOF && + relpath.c + EOF + git diff --cached --name-only >actual && + test_cmp expect actual +' + test_expect_success 'add -p does not expand argument lists' ' git reset --hard && -- 2.12.2 ``` ## Brandon Williams, 2017-04-04 16:39 Subject: Re: [PATCH v2] pathspec: honor `PATHSPEC_PREFIX_ORIGIN` with empty prefix Message-ID: <20170404163909.GA189807@google.com> URL: https://gitlist.dev/e/20170404163909.GA189807%40google.com In-Reply-To: <4650e1396a5ad006b2b4f355685e3c59efd554d6.1491297393.git.ps@pks.im> ``` On 04/04, Patrick Steinhardt wrote: > Previous to commit 5d8f084a5 (pathspec: simpler logic to prefix original > pathspec elements, 2017-01-04), we were always using the computed > `match` variable to perform pathspec matching whenever > `PATHSPEC_PREFIX_ORIGIN` is set. This is for example useful when passing > the parsed pathspecs to other commands, as the computed `match` may > contain a pathspec relative to the repository root. The commit changed > this logic to only do so when we do have an actual prefix and when > literal pathspecs are deactivated. > > But this change may actually break some commands which expect passed > pathspecs to be relative to the repository root. One such case is `git > add --patch`, which now fails when using relative paths from a > subdirectory. For example if executing "git add -p ../foo.c" in a > subdirectory, the `git-add--interactive` command will directly pass > "../foo.c" to `git-ls-files`. As ls-files is executed at the > repository's root, the command will notice that "../foo.c" is outside > the repository and fail. > > Fix the issue by again using the computed `match` variable when > `PATHSPEC_PREFIX_ORIGIN` is set and global literal pathspecs are > deactivated. Note that in contrast to previous behavior, we will now > always call `prefix_magic` regardless of whether a prefix is actually > set. But this is the right thing to do: when the `match` variable has > been resolved to the repository's root, it will be set to an empty > string. When passing the empty string directly to other commands, it > will result in a warning regarding deprecated empty pathspecs. By always > adding the prefix magic, we will end up with at least the string > ":(prefix:0)" and thus avoid the warning. > > Signed-off-by: Patrick Steinhardt > --- > > This is the second version of [1]. It fixes a bug catched by > Brandon when the pathspec is resolved to the empty string and > improves the test a bit to actually catch this issue. This version looks good to me. Thanks for fixing that small issue! -- Brandon Williams ``` ## Duy Nguyen, 2017-04-05 13:46 Subject: Re: [PATCH v2] pathspec: honor `PATHSPEC_PREFIX_ORIGIN` with empty prefix Message-ID: URL: https://gitlist.dev/e/CACsJy8AeZWmGuvEqoA4y6rM-kiz6_7r31Q4kqor7%2BsHSwgZ1sQ%40mail.gmail.com In-Reply-To: <20170404163909.GA189807@google.com> ``` On Tue, Apr 4, 2017 at 11:39 PM, Brandon Williams wrote: >> This is the second version of [1]. It fixes a bug catched by >> Brandon when the pathspec is resolved to the empty string and >> improves the test a bit to actually catch this issue. > > This version looks good to me. Thanks for fixing that small issue! Looks good to me too (I didn't see this when I read v1, too much mails) -- Duy ``` ## Patrick Steinhardt, 2017-04-12 08:22 Subject: Re: [PATCH v2] pathspec: honor `PATHSPEC_PREFIX_ORIGIN` with empty prefix Message-ID: <20170412082205.GA448@pks-xps> URL: https://gitlist.dev/e/20170412082205.GA448%40pks-xps In-Reply-To: <4650e1396a5ad006b2b4f355685e3c59efd554d6.1491297393.git.ps@pks.im> ``` On Tue, Apr 04, 2017 at 11:16:56AM +0200, Patrick Steinhardt wrote: > Previous to commit 5d8f084a5 (pathspec: simpler logic to prefix original > pathspec elements, 2017-01-04), we were always using the computed > `match` variable to perform pathspec matching whenever > `PATHSPEC_PREFIX_ORIGIN` is set. This is for example useful when passing > the parsed pathspecs to other commands, as the computed `match` may > contain a pathspec relative to the repository root. The commit changed > this logic to only do so when we do have an actual prefix and when > literal pathspecs are deactivated. > > But this change may actually break some commands which expect passed > pathspecs to be relative to the repository root. One such case is `git > add --patch`, which now fails when using relative paths from a > subdirectory. For example if executing "git add -p ../foo.c" in a > subdirectory, the `git-add--interactive` command will directly pass > "../foo.c" to `git-ls-files`. As ls-files is executed at the > repository's root, the command will notice that "../foo.c" is outside > the repository and fail. > > Fix the issue by again using the computed `match` variable when > `PATHSPEC_PREFIX_ORIGIN` is set and global literal pathspecs are > deactivated. Note that in contrast to previous behavior, we will now > always call `prefix_magic` regardless of whether a prefix is actually > set. But this is the right thing to do: when the `match` variable has > been resolved to the repository's root, it will be set to an empty > string. When passing the empty string directly to other commands, it > will result in a warning regarding deprecated empty pathspecs. By always > adding the prefix magic, we will end up with at least the string > ":(prefix:0)" and thus avoid the warning. > > Signed-off-by: Patrick Steinhardt > --- Just a short reminder on this patch, as I haven't seen it or v1 being picked up by the What's Cooking reports. Am I simply being too eager or was this an oversight? Thanks Patrick > > This is the second version of [1]. It fixes a bug catched by > Brandon when the pathspec is resolved to the empty string and > improves the test a bit to actually catch this issue. > > [1]: http://public-inbox.org/git/1556910880cfce391bdca2d8f0cbcb8c71371691.1491206540.git.ps@pks.im/T/#u > > > pathspec.c | 2 +- > t/t3701-add-interactive.sh | 22 ++++++++++++++++++++++ > 2 files changed, 23 insertions(+), 1 deletion(-) > > diff --git a/pathspec.c b/pathspec.c > index 303efda83..3079a817a 100644 > --- a/pathspec.c > +++ b/pathspec.c > @@ -505,7 +505,7 @@ static void init_pathspec_item(struct pathspec_item *item, unsigned flags, > * original. Useful for passing to another command. > */ > if ((flags & PATHSPEC_PREFIX_ORIGIN) && > - prefixlen && !get_literal_global()) { > + !get_literal_global()) { > struct strbuf sb = STRBUF_INIT; > > /* Preserve the actual prefix length of each pattern */ > diff --git a/t/t3701-add-interactive.sh b/t/t3701-add-interactive.sh > index f9528fa00..2ecb43a61 100755 > --- a/t/t3701-add-interactive.sh > +++ b/t/t3701-add-interactive.sh > @@ -436,6 +436,28 @@ test_expect_success 'add -p handles globs' ' > test_cmp expect actual > ' > > +test_expect_success 'add -p handles relative paths' ' > + git reset --hard && > + > + echo base >relpath.c && > + git add "*.c" && > + git commit -m relpath && > + > + echo change >relpath.c && > + mkdir -p subdir && > + git -C subdir add -p .. 2>error <<-\EOF && > + y > + EOF > + > + test_must_be_empty error && > + > + cat >expect <<-\EOF && > + relpath.c > + EOF > + git diff --cached --name-only >actual && > + test_cmp expect actual > +' > + > test_expect_success 'add -p does not expand argument lists' ' > git reset --hard && > > -- > 2.12.2 ``` ## Junio C Hamano, 2017-04-12 08:47 Subject: Re: [PATCH v2] pathspec: honor `PATHSPEC_PREFIX_ORIGIN` with empty prefix Message-ID: URL: https://gitlist.dev/e/xmqqr30yovvs.fsf%40gitster.mtv.corp.google.com In-Reply-To: <20170412082205.GA448@pks-xps> ``` Patrick Steinhardt writes: > Just a short reminder on this patch, as I haven't seen it or > v1 being picked up by the What's Cooking reports. Am I simply > being too eager or was this an oversight? I have been offline and will be for a few more days; I may occasionally pop in to give design level guidance on topics, but won't be reading patches carefully enough to queue a new topic to my tree. New topics won't be added to my tree until next week the earliest. ```