# [PATCH] pathspec: warn on empty strings as pathspec 5 messages from 2016-06-22 to 2016-06-24. Participants: Emily Xie, Junio C Hamano. Thread: https://gitlist.dev/t/42695 ## Emily Xie, 2016-06-22 23:00 Subject: [PATCH] pathspec: warn on empty strings as pathspec Message-ID: <20160622230024.24034-1-emilyxxie@gmail.com> URL: https://gitlist.dev/e/20160622230024.24034-1-emilyxxie%40gmail.com ``` Currently, passing an empty string as a pathspec results in a match on all paths. This is because a pathspec match is a leading substring match that honors directory boundaries. So, just as pathspec "dir/" (or "dir") matches "dir/file", "" matches "file". However, this causes problems. Namely, a user's carelessly-written script could accidentally assign an empty string to a variable that then gets passed to a Git command invocation, e.g.: git rm -r "$path" git add "$path" which would unintentionally affect all paths in the current directory. The fix for this issue requires a two-step approach. As there may be existing scripts that knowingly use empty strings in this manner, the first step simply invokes a warning that (1) declares using empty strings to match all paths will become invalid and (2) asks the user to use "." if their intent was to match all. For step two, a follow-up patch several release cycles later will remove the warnings and throw an error instead. This patch is the first step. Signed-off-by: Emily Xie Reported-by: David Turner Mentored-by: Michail Denchev Thanks-to: Sarah Sharp and James Sharp --- pathspec.c | 11 +++++++++-- t/t3600-rm.sh | 5 +++++ t/t3700-add.sh | 5 +++++ 3 files changed, 19 insertions(+), 2 deletions(-) diff --git a/pathspec.c b/pathspec.c index c9e9b6c..02aa691 100644 --- a/pathspec.c +++ b/pathspec.c @@ -364,7 +364,7 @@ void parse_pathspec(struct pathspec *pathspec, { struct pathspec_item *item; const char *entry = argv ? *argv : NULL; - int i, n, prefixlen, nr_exclude = 0; + int i, n, prefixlen, warn_empty_string, nr_exclude = 0; memset(pathspec, 0, sizeof(*pathspec)); @@ -402,8 +402,15 @@ void parse_pathspec(struct pathspec *pathspec, } n = 0; - while (argv[n]) + warn_empty_string = 1; + while (argv[n]) { + if (*argv[n] == '\0' && warn_empty_string) { + warning(_("empty strings as pathspecs will be made invalid in upcoming releases. " + "please use . instead if you meant to match all paths")); + warn_empty_string = 0; + } n++; + } pathspec->nr = n; ALLOC_ARRAY(pathspec->items, n); diff --git a/t/t3600-rm.sh b/t/t3600-rm.sh index d046d98..14f0edc 100755 --- a/t/t3600-rm.sh +++ b/t/t3600-rm.sh @@ -881,4 +881,9 @@ test_expect_success 'rm files with two different errors' ' test_i18ncmp expect actual ' +test_expect_success 'rm empty string should invoke warning' ' + git rm -rf "" 2>output && + test_i18ngrep "warning: empty strings" output +' + test_done diff --git a/t/t3700-add.sh b/t/t3700-add.sh index f14a665..05379d0 100755 --- a/t/t3700-add.sh +++ b/t/t3700-add.sh @@ -332,4 +332,9 @@ test_expect_success 'git add --dry-run --ignore-missing of non-existing file out test_i18ncmp expect.err actual.err ' +test_expect_success 'git add empty string should invoke warning' ' + git add "" 2>output && + test_i18ngrep "warning: empty strings" output +' + test_done -- 2.8.4 ``` ## Junio C Hamano, 2016-06-22 23:12 Subject: Re: [PATCH] pathspec: warn on empty strings as pathspec Message-ID: URL: https://gitlist.dev/e/xmqqmvmckebv.fsf%40gitster.mtv.corp.google.com In-Reply-To: <20160622230024.24034-1-emilyxxie@gmail.com> ``` Emily Xie writes: > The fix for this issue requires a two-step approach. As > there may be existing scripts that knowingly use empty > strings in this manner, the first step simply invokes a > warning that (1) declares using empty strings to > match all paths will become invalid and (2) asks the user > to use "." if their intent was to match all. > > For step two, a follow-up patch several release cycles > later will remove the warnings and throw an error instead. > > This patch is the first step. > > Signed-off-by: Emily Xie > Reported-by: David Turner > Mentored-by: Michail Denchev > Thanks-to: Sarah Sharp and James Sharp > --- Looks sensible. I personally do not think warn-only-once matters in practice, but now you have implemented it, it is an additional nice touch ;-) Will queue. Thanks. ``` ## Emily Xie, 2016-06-22 23:48 Subject: Re: [PATCH] pathspec: warn on empty strings as pathspec Message-ID: URL: https://gitlist.dev/e/CAAin2AQR3yP7X%3DNLQ%3D9%2B0QGJZMCN7%3DxbpZ%3DH7r56RvvBfLJSEg%40mail.gmail.com In-Reply-To: ``` Awesome. Thanks, Junio---this is exciting! As for step 2: do you have a good sense on the timing? Around how long should I wait to submit the patch for it? Otherwise, let me know if there is anything else that I need to do! - Emily ``` ## Junio C Hamano, 2016-06-23 06:17 Subject: Re: [PATCH] pathspec: warn on empty strings as pathspec Message-ID: URL: https://gitlist.dev/e/xmqqh9ckjuon.fsf%40gitster.mtv.corp.google.com In-Reply-To: ``` Emily Xie writes: > Awesome. Thanks, Junio---this is exciting! No, thank _you_. > As for step 2: do you have a good sense on the timing? Around how long > should I wait to submit the patch for it? Not so fast. We'll wait for others to comment first. I am queuing this change but that does not mean anything more than that I am not outright rejecting the idea. It is possible that others may assess the cost of having to do the two-step migration differently, and the list concensus may end up being "if it hurts, don't pass an empty string", i.e. we'd better without this patch or subsequent second step. If that happens, the first step dies without hitting 'next'. I'd say we'd wait at least for a week. Otherwise, if the change is received favourably, we'll merge it to 'next', do the same waiting for comments. Repeat the same and then merge to 'master'. After it hits the next feature release (probably at around the end of August), the change will be seen by wider general public, and at that point we may see strong opposition from them with a good argument neither of us thought of why we shouldn't do this change, in which case we might have to revise the plan and scrap the whole thing. So, we'll wait and see. ``` ## Emily Xie, 2016-06-24 18:21 Subject: Re: [PATCH] pathspec: warn on empty strings as pathspec Message-ID: URL: https://gitlist.dev/e/CAAin2ASggO1ZUqT_nfokkYL40vSDSdPtAVq4H2Z9A7K633xq3Q%40mail.gmail.com In-Reply-To: ``` I suppose I got ahead of myself there. :-) Thanks for the clarification on the process. On Thu, Jun 23, 2016 at 2:17 AM, Junio C Hamano wrote: > Emily Xie writes: > >> Awesome. Thanks, Junio---this is exciting! > > No, thank _you_. > >> As for step 2: do you have a good sense on the timing? Around how long >> should I wait to submit the patch for it? > > Not so fast. We'll wait for others to comment first. > > I am queuing this change but that does not mean anything more than > that I am not outright rejecting the idea. > > It is possible that others may assess the cost of having to do the > two-step migration differently, and the list concensus may end up > being "if it hurts, don't pass an empty string", i.e. we'd better > without this patch or subsequent second step. If that happens, the > first step dies without hitting 'next'. I'd say we'd wait at least > for a week. > > Otherwise, if the change is received favourably, we'll merge it to > 'next', do the same waiting for comments. Repeat the same and then > merge to 'master'. After it hits the next feature release (probably > at around the end of August), the change will be seen by wider > general public, and at that point we may see strong opposition from > them with a good argument neither of us thought of why we shouldn't > do this change, in which case we might have to revise the plan and > scrap the whole thing. > > So, we'll wait and see. > -- Emily Xie xie-emily.com 207.399.6370 ```