# [PATCH] fsck: avoid looking at NULL blob->object 17 messages from 2018-06-09 to 2018-06-12. Participants: Jeff King, Eric Sunshine, Duy Nguyen, Martin Ågren, SZEDER Gábor. Thread: https://gitlist.dev/t/48680 ## Jeff King, 2018-06-09 08:32 Subject: [PATCH] fsck: avoid looking at NULL blob->object Message-ID: <20180609083159.GA6158@sigill.intra.peff.net> URL: https://gitlist.dev/e/20180609083159.GA6158%40sigill.intra.peff.net ``` Commit 159e7b080b (fsck: detect gitmodules files, 2018-05-02) taught fsck to look at the content of .gitmodules files. If the object turns out not to be a blob at all, we just complain and punt on checking the content. And since this was such an obvious and trivial code path, I didn't even bother to add a test. Except it _does_ do one non-trivial thing, which is call the report() function, which wants us to pass a pointer to a "struct object". Which we don't have (we have only a "struct object_id"). So we erroneously passed the NULL object, which ends up segfaulting. It seems like we could refactor report() to just take the object_id itself. But we pass the object pointer along to a callback function, and indeed this ends up in builtin/fsck.c's objreport() which does want to look at other parts of the object (like the type). So instead, let's just use lookup_unknown_object() to get the real "struct object", and pass that. Signed-off-by: Jeff King --- The problem is in v2.17.1, and of course the v2.18 release candidates. fsck.c | 3 ++- t/t7415-submodule-names.sh | 18 ++++++++++++++++++ 2 files changed, 20 insertions(+), 1 deletion(-) diff --git a/fsck.c b/fsck.c index bcae2c30e6..48e7e36869 100644 --- a/fsck.c +++ b/fsck.c @@ -1036,7 +1036,8 @@ int fsck_finish(struct fsck_options *options) blob = lookup_blob(oid); if (!blob) { - ret |= report(options, &blob->object, + struct object *obj = lookup_unknown_object(oid->hash); + ret |= report(options, obj, FSCK_MSG_GITMODULES_BLOB, "non-blob found at .gitmodules"); continue; diff --git a/t/t7415-submodule-names.sh b/t/t7415-submodule-names.sh index a770d92a55..63c767bf99 100755 --- a/t/t7415-submodule-names.sh +++ b/t/t7415-submodule-names.sh @@ -151,4 +151,22 @@ test_expect_success 'fsck detects symlinked .gitmodules file' ' ) ' +test_expect_success 'fsck detects non-blob .gitmodules' ' + git init non-blob && + ( + cd non-blob && + + # As above, make the funny directly to avoid index restrictions. + mkdir subdir && + cp ../.gitmodules subdir/file && + git add subdir/file && + git commit -m ok && + tree=$(git ls-tree HEAD | sed s/subdir/.gitmodules/ | git mktree) && + commit=$(git commit-tree $tree) && + + test_must_fail git fsck 2>output && + grep gitmodulesBlob output + ) +' + test_done -- 2.18.0.rc1.446.g4486251e51 ``` ## Eric Sunshine, 2018-06-09 08:38 Subject: Re: [PATCH] fsck: avoid looking at NULL blob->object Message-ID: URL: https://gitlist.dev/e/CAPig%2BcTgCD5%3D96XG%3DZ5FwOsPbN409DxzAfPy0p%3DwnoLywu%2B%2Bdw%40mail.gmail.com In-Reply-To: <20180609083159.GA6158@sigill.intra.peff.net> ``` On Sat, Jun 9, 2018 at 4:32 AM, Jeff King wrote: > Commit 159e7b080b (fsck: detect gitmodules files, > 2018-05-02) taught fsck to look at the content of > .gitmodules files. If the object turns out not to be a blob > at all, we just complain and punt on checking the content. > And since this was such an obvious and trivial code path, I > didn't even bother to add a test. > [...] > Signed-off-by: Jeff King > --- > diff --git a/t/t7415-submodule-names.sh b/t/t7415-submodule-names.sh > @@ -151,4 +151,22 @@ test_expect_success 'fsck detects symlinked .gitmodules file' ' > +test_expect_success 'fsck detects non-blob .gitmodules' ' > + git init non-blob && > + ( > + cd non-blob && > + > + # As above, make the funny directly to avoid index restrictions. Is there a word missing after "funny"? > + mkdir subdir && > + cp ../.gitmodules subdir/file && > + git add subdir/file && > + git commit -m ok && > + tree=$(git ls-tree HEAD | sed s/subdir/.gitmodules/ | git mktree) && > + commit=$(git commit-tree $tree) && I see that this is just mirroring the preceding test, but do you need to assign to variable 'commit' which is never consulted by anything later in the test? > + test_must_fail git fsck 2>output && > + grep gitmodulesBlob output > + ) > +' ``` ## Duy Nguyen, 2018-06-09 08:45 Subject: Re: [PATCH] fsck: avoid looking at NULL blob->object Message-ID: URL: https://gitlist.dev/e/CACsJy8Cpd_0CR9F_xyQfcw0j8ceOf777KcWXLHuxY1NxmRnXqw%40mail.gmail.com In-Reply-To: <20180609083159.GA6158@sigill.intra.peff.net> ``` On Sat, Jun 9, 2018 at 10:34 AM Jeff King wrote: > > Commit 159e7b080b (fsck: detect gitmodules files, > 2018-05-02) taught fsck to look at the content of > .gitmodules files. If the object turns out not to be a blob > at all, we just complain and punt on checking the content. > And since this was such an obvious and trivial code path, I > didn't even bother to add a test. > > Except it _does_ do one non-trivial thing, which is call the > report() function, which wants us to pass a pointer to a > "struct object". Which we don't have (we have only a "struct > object_id"). So we erroneously passed the NULL object, which > ends up segfaulting. > > It seems like we could refactor report() to just take the > object_id itself. But we pass the object pointer along to > a callback function, and indeed this ends up in > builtin/fsck.c's objreport() which does want to look at > other parts of the object (like the type). And objreport() can handle OBJ_NONE well, which is the type given by lookup_unknown_object(). So yeah this looks good. > So instead, let's just use lookup_unknown_object() to get > the real "struct object", and pass that. -- Duy ``` ## Martin Ågren, 2018-06-09 08:50 Subject: Re: [PATCH] fsck: avoid looking at NULL blob->object Message-ID: URL: https://gitlist.dev/e/CAN0heSqHRpOerHV8tgA-ntU6QvYQMxgrXPCPsrVcHLKUSwANXQ%40mail.gmail.com In-Reply-To: <20180609083159.GA6158@sigill.intra.peff.net> ``` On 9 June 2018 at 10:32, Jeff King wrote: > Except it _does_ do one non-trivial thing, which is call the > report() function, which wants us to pass a pointer to a > "struct object". Which we don't have (we have only a "struct > object_id"). So we erroneously passed the NULL object, which s/passed/dereferenced/? Probably doesn't affect the fix though. > ends up segfaulting. > blob = lookup_blob(oid); > if (!blob) { > - ret |= report(options, &blob->object, > + struct object *obj = lookup_unknown_object(oid->hash); > + ret |= report(options, obj, > FSCK_MSG_GITMODULES_BLOB, > "non-blob found at .gitmodules"); ``` ## Jeff King, 2018-06-09 09:19 Subject: Re: [PATCH] fsck: avoid looking at NULL blob->object Message-ID: <20180609091945.GA6817@sigill.intra.peff.net> URL: https://gitlist.dev/e/20180609091945.GA6817%40sigill.intra.peff.net In-Reply-To: ``` On Sat, Jun 09, 2018 at 04:38:54AM -0400, Eric Sunshine wrote: > On Sat, Jun 9, 2018 at 4:32 AM, Jeff King wrote: > > Commit 159e7b080b (fsck: detect gitmodules files, > > 2018-05-02) taught fsck to look at the content of > > .gitmodules files. If the object turns out not to be a blob > > at all, we just complain and punt on checking the content. > > And since this was such an obvious and trivial code path, I > > didn't even bother to add a test. > > [...] > > Signed-off-by: Jeff King > > --- > > diff --git a/t/t7415-submodule-names.sh b/t/t7415-submodule-names.sh > > @@ -151,4 +151,22 @@ test_expect_success 'fsck detects symlinked .gitmodules file' ' > > +test_expect_success 'fsck detects non-blob .gitmodules' ' > > + git init non-blob && > > + ( > > + cd non-blob && > > + > > + # As above, make the funny directly to avoid index restrictions. > > Is there a word missing after "funny"? Oops, should be "funny tree" (that's what I get for trying to wordsmith it at the last minute). > > + mkdir subdir && > > + cp ../.gitmodules subdir/file && > > + git add subdir/file && > > + git commit -m ok && > > + tree=$(git ls-tree HEAD | sed s/subdir/.gitmodules/ | git mktree) && > > + commit=$(git commit-tree $tree) && > > I see that this is just mirroring the preceding test, but do you need > to assign to variable 'commit' which is never consulted by anything > later in the test? No (nor above). I think originally I had planned to points refs to these commits, but it isn't necessary for fsck. In fact, in the final form of the patches, we do not even need the commit at all, since we will complain about .gitmodules in _any_ tree (early versions actually did an extra pass to find which trees were root trees). -Peff ``` ## Jeff King, 2018-06-09 09:20 Subject: Re: [PATCH] fsck: avoid looking at NULL blob->object Message-ID: <20180609092056.GB6817@sigill.intra.peff.net> URL: https://gitlist.dev/e/20180609092056.GB6817%40sigill.intra.peff.net In-Reply-To: ``` On Sat, Jun 09, 2018 at 10:45:15AM +0200, Duy Nguyen wrote: > > It seems like we could refactor report() to just take the > > object_id itself. But we pass the object pointer along to > > a callback function, and indeed this ends up in > > builtin/fsck.c's objreport() which does want to look at > > other parts of the object (like the type). > > And objreport() can handle OBJ_NONE well, which is the type given by > lookup_unknown_object(). So yeah this looks good. Actually, we know that we have some "real" type, because otherwise lookup_blob() would not have returned NULL. In fact, we could actually use lookup_object() here to find that object, knowing that it exists in the hash. But IMHO that would be depending too much on how lookup_blob() works. :) -Peff ``` ## Jeff King, 2018-06-09 09:21 Subject: Re: [PATCH] fsck: avoid looking at NULL blob->object Message-ID: <20180609092135.GC6817@sigill.intra.peff.net> URL: https://gitlist.dev/e/20180609092135.GC6817%40sigill.intra.peff.net In-Reply-To: ``` On Sat, Jun 09, 2018 at 10:50:36AM +0200, Martin Ågren wrote: > On 9 June 2018 at 10:32, Jeff King wrote: > > Except it _does_ do one non-trivial thing, which is call the > > report() function, which wants us to pass a pointer to a > > "struct object". Which we don't have (we have only a "struct > > object_id"). So we erroneously passed the NULL object, which > > s/passed/dereferenced/? Probably doesn't affect the fix though. Well, we passed it, and then that function dereferenced it. :) I'm going to re-roll for the minor bits that Eric pointed out, so I'll try to word this better. -Peff ``` ## Jeff King, 2018-06-09 09:30 Subject: [PATCH v2 0/2] .gitmodules fsck cleanups Message-ID: <20180609093028.GA8864@sigill.intra.peff.net> URL: https://gitlist.dev/e/20180609093028.GA8864%40sigill.intra.peff.net In-Reply-To: <20180609083159.GA6158@sigill.intra.peff.net> ``` Here's a v2 which fixes the comment typo and drops the unnecessary commit creation in the tests. I've added a preparatory patch to do the same cleanup in the existing test for consistency. As before, these should apply on 'maint'. [1/2]: t7415: don't bother creating commit for symlink test [2/2]: fsck: avoid looking at NULL blob->object fsck.c | 3 ++- t/t7415-submodule-names.sh | 19 ++++++++++++++++++- 2 files changed, 20 insertions(+), 2 deletions(-) -Peff ``` ## Jeff King, 2018-06-09 09:31 Subject: [PATCH v2 1/2] t7415: don't bother creating commit for symlink test Message-ID: <20180609093118.GA8942@sigill.intra.peff.net> URL: https://gitlist.dev/e/20180609093118.GA8942%40sigill.intra.peff.net In-Reply-To: <20180609093028.GA8864@sigill.intra.peff.net> ``` Early versions of the fsck .gitmodules detection code actually required a tree to be at the root of a commit for it to be checked for .gitmodules. What we ended up with in 159e7b080b (fsck: detect gitmodules files, 2018-05-02), though, finds a .gitmodules file in _any_ tree (see that commit for more discussion). As a result, there's no need to create a commit in our tests. Let's drop it in the name of simplicity. Signed-off-by: Jeff King --- t/t7415-submodule-names.sh | 1 - 1 file changed, 1 deletion(-) diff --git a/t/t7415-submodule-names.sh b/t/t7415-submodule-names.sh index a770d92a55..e2aae587ae 100755 --- a/t/t7415-submodule-names.sh +++ b/t/t7415-submodule-names.sh @@ -141,7 +141,6 @@ test_expect_success 'fsck detects symlinked .gitmodules file' ' printf "120000 blob $target\t.gitmodules\n" } | git mktree ) && - commit=$(git commit-tree $tree) && # Check not only that we fail, but that it is due to the # symlink detector; this grep string comes from the config -- 2.18.0.rc1.446.g4486251e51 ``` ## Jeff King, 2018-06-09 09:31 Subject: [PATCH v2 2/2] fsck: avoid looking at NULL blob->object Message-ID: <20180609093141.GB8942@sigill.intra.peff.net> URL: https://gitlist.dev/e/20180609093141.GB8942%40sigill.intra.peff.net In-Reply-To: <20180609093028.GA8864@sigill.intra.peff.net> ``` Commit 159e7b080b (fsck: detect gitmodules files, 2018-05-02) taught fsck to look at the content of .gitmodules files. If the object turns out not to be a blob at all, we just complain and punt on checking the content. And since this was such an obvious and trivial code path, I didn't even bother to add a test. Except it _does_ do one non-trivial thing, which is call the report() function, which wants us to pass a pointer to a "struct object". Which we don't have (we have only a "struct object_id"). So we erroneously pass a NULL object to report(), which gets dereferenced and causes a segfault. It seems like we could refactor report() to just take the object_id itself. But we pass the object pointer along to a callback function, and indeed this ends up in builtin/fsck.c's objreport() which does want to look at other parts of the object (like the type). So instead, let's just use lookup_unknown_object() to get the real "struct object", and pass that. Signed-off-by: Jeff King --- fsck.c | 3 ++- t/t7415-submodule-names.sh | 18 ++++++++++++++++++ 2 files changed, 20 insertions(+), 1 deletion(-) diff --git a/fsck.c b/fsck.c index bcae2c30e6..48e7e36869 100644 --- a/fsck.c +++ b/fsck.c @@ -1036,7 +1036,8 @@ int fsck_finish(struct fsck_options *options) blob = lookup_blob(oid); if (!blob) { - ret |= report(options, &blob->object, + struct object *obj = lookup_unknown_object(oid->hash); + ret |= report(options, obj, FSCK_MSG_GITMODULES_BLOB, "non-blob found at .gitmodules"); continue; diff --git a/t/t7415-submodule-names.sh b/t/t7415-submodule-names.sh index e2aae587ae..d7868c11fa 100755 --- a/t/t7415-submodule-names.sh +++ b/t/t7415-submodule-names.sh @@ -150,4 +150,22 @@ test_expect_success 'fsck detects symlinked .gitmodules file' ' ) ' +test_expect_success 'fsck detects non-blob .gitmodules' ' + git init non-blob && + ( + cd non-blob && + + # As above, make the funny tree directly to avoid index + # restrictions. + mkdir subdir && + cp ../.gitmodules subdir/file && + git add subdir/file && + git commit -m ok && + tree=$(git ls-tree HEAD | sed s/subdir/.gitmodules/ | git mktree) && + + test_must_fail git fsck 2>output && + grep gitmodulesBlob output + ) +' + test_done -- 2.18.0.rc1.446.g4486251e51 ``` ## SZEDER Gábor, 2018-06-09 09:46 Subject: Re: [PATCH v2 1/2] t7415: don't bother creating commit for symlink test Message-ID: <20180609094613.17675-1-szeder.dev@gmail.com> URL: https://gitlist.dev/e/20180609094613.17675-1-szeder.dev%40gmail.com In-Reply-To: <20180609093118.GA8942@sigill.intra.peff.net> ``` > As a result, there's no need to create a commit in our > tests. Let's drop it in the name of simplicity. > > Signed-off-by: Jeff King > --- > t/t7415-submodule-names.sh | 1 - > 1 file changed, 1 deletion(-) > > diff --git a/t/t7415-submodule-names.sh b/t/t7415-submodule-names.sh > index a770d92a55..e2aae587ae 100755 > --- a/t/t7415-submodule-names.sh > +++ b/t/t7415-submodule-names.sh > @@ -141,7 +141,6 @@ test_expect_success 'fsck detects symlinked .gitmodules file' ' I add few more lines of context here: tree=$( { printf "100644 blob $content\t$tricky\n" && > printf "120000 blob $target\t.gitmodules\n" > } | git mktree > ) && > - commit=$(git commit-tree $tree) && This was the only case where that $tree variable was used, so perhaps that can go away as well, in the name of even more simplicity? > > # Check not only that we fail, but that it is due to the > # symlink detector; this grep string comes from the config > -- > 2.18.0.rc1.446.g4486251e51 > > ``` ## SZEDER Gábor, 2018-06-09 09:50 Subject: Re: [PATCH v2 2/2] fsck: avoid looking at NULL blob->object Message-ID: <20180609095014.17972-1-szeder.dev@gmail.com> URL: https://gitlist.dev/e/20180609095014.17972-1-szeder.dev%40gmail.com In-Reply-To: <20180609093141.GB8942@sigill.intra.peff.net> ``` > +test_expect_success 'fsck detects non-blob .gitmodules' ' > + git init non-blob && > + ( > + cd non-blob && > + > + # As above, make the funny tree directly to avoid index > + # restrictions. > + mkdir subdir && > + cp ../.gitmodules subdir/file && > + git add subdir/file && > + git commit -m ok && > + tree=$(git ls-tree HEAD | sed s/subdir/.gitmodules/ | git mktree) && And $tree won't be used here, either. > + > + test_must_fail git fsck 2>output && > + grep gitmodulesBlob output > + ) > +' > + > test_done > -- > 2.18.0.rc1.446.g4486251e51 > ``` ## Martin Ågren, 2018-06-09 13:44 Subject: Re: [PATCH] fsck: avoid looking at NULL blob->object Message-ID: URL: https://gitlist.dev/e/CAN0heSqkcZ5KBpaq7R45A7wPxQ3F%3D6PKeHqwAQztdJyz31Vdwg%40mail.gmail.com In-Reply-To: <20180609092135.GC6817@sigill.intra.peff.net> ``` On 9 June 2018 at 11:21, Jeff King wrote: > On Sat, Jun 09, 2018 at 10:50:36AM +0200, Martin Ågren wrote: > >> On 9 June 2018 at 10:32, Jeff King wrote: >> > Except it _does_ do one non-trivial thing, which is call the >> > report() function, which wants us to pass a pointer to a >> > "struct object". Which we don't have (we have only a "struct >> > object_id"). So we erroneously passed the NULL object, which >> >> s/passed/dereferenced/? Probably doesn't affect the fix though. > > Well, we passed it, and then that function dereferenced it. :) > > I'm going to re-roll for the minor bits that Eric pointed out, so I'll > try to word this better. My bad. I somehow thought we get into trouble already before we call `report()`. Well, we do, since we have undefined behavior. But for all practical purposes `&blob->object` and `blob` are the same (NULL-)pointer so we only crash after we call `report()`. Anyway, obviously no need to do anything about this in a v3. Martin ``` ## Jeff King, 2018-06-11 08:35 Subject: [PATCH v3 0/2] .gitmodules fsck cleanups Message-ID: <20180611083515.GA1815@sigill.intra.peff.net> URL: https://gitlist.dev/e/20180611083515.GA1815%40sigill.intra.peff.net In-Reply-To: <20180609094613.17675-1-szeder.dev@gmail.com> ``` On Sat, Jun 09, 2018 at 11:46:13AM +0200, SZEDER Gábor wrote: > I add few more lines of context here: > > tree=$( > { > printf "100644 blob $content\t$tricky\n" && > > > printf "120000 blob $target\t.gitmodules\n" > > } | git mktree > > ) && > > - commit=$(git commit-tree $tree) && > > This was the only case where that $tree variable was used, so perhaps > that can go away as well, in the name of even more simplicity? Yep, that is simpler. Here's a v3 just dropping those $tree variables, and adjusting the commit message as appropriate. [1/2]: t7415: don't bother creating commit for symlink test [2/2]: fsck: avoid looking at NULL blob->object fsck.c | 3 ++- t/t7415-submodule-names.sh | 29 ++++++++++++++++++++++------- 2 files changed, 24 insertions(+), 8 deletions(-) -Peff ``` ## Jeff King, 2018-06-11 08:35 Subject: [PATCH v3 1/2] t7415: don't bother creating commit for symlink test Message-ID: <20180611083540.GA16131@sigill.intra.peff.net> URL: https://gitlist.dev/e/20180611083540.GA16131%40sigill.intra.peff.net In-Reply-To: <20180611083515.GA1815@sigill.intra.peff.net> ``` Early versions of the fsck .gitmodules detection code actually required a tree to be at the root of a commit for it to be checked for .gitmodules. What we ended up with in 159e7b080b (fsck: detect gitmodules files, 2018-05-02), though, finds a .gitmodules file in _any_ tree (see that commit for more discussion). As a result, there's no need to create a commit in our tests. Let's drop it in the name of simplicity. And since that was the only thing referencing $tree, we can pull our tree creation out of a command substitution. Signed-off-by: Jeff King --- t/t7415-submodule-names.sh | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-) diff --git a/t/t7415-submodule-names.sh b/t/t7415-submodule-names.sh index a770d92a55..541bd81684 100755 --- a/t/t7415-submodule-names.sh +++ b/t/t7415-submodule-names.sh @@ -135,13 +135,10 @@ test_expect_success 'fsck detects symlinked .gitmodules file' ' tricky="[foo]bar=true" && content=$(git hash-object -w ../.gitmodules) && target=$(printf "$tricky" | git hash-object -w --stdin) && - tree=$( - { - printf "100644 blob $content\t$tricky\n" && - printf "120000 blob $target\t.gitmodules\n" - } | git mktree - ) && - commit=$(git commit-tree $tree) && + { + printf "100644 blob $content\t$tricky\n" && + printf "120000 blob $target\t.gitmodules\n" + } | git mktree && # Check not only that we fail, but that it is due to the # symlink detector; this grep string comes from the config -- 2.18.0.rc1.446.g4486251e51 ``` ## Jeff King, 2018-06-11 08:35 Subject: [PATCH v3 2/2] fsck: avoid looking at NULL blob->object Message-ID: <20180611083544.GB16131@sigill.intra.peff.net> URL: https://gitlist.dev/e/20180611083544.GB16131%40sigill.intra.peff.net In-Reply-To: <20180611083515.GA1815@sigill.intra.peff.net> ``` Commit 159e7b080b (fsck: detect gitmodules files, 2018-05-02) taught fsck to look at the content of .gitmodules files. If the object turns out not to be a blob at all, we just complain and punt on checking the content. And since this was such an obvious and trivial code path, I didn't even bother to add a test. Except it _does_ do one non-trivial thing, which is call the report() function, which wants us to pass a pointer to a "struct object". Which we don't have (we have only a "struct object_id"). So we erroneously pass a NULL object to report(), which gets dereferenced and causes a segfault. It seems like we could refactor report() to just take the object_id itself. But we pass the object pointer along to a callback function, and indeed this ends up in builtin/fsck.c's objreport() which does want to look at other parts of the object (like the type). So instead, let's just use lookup_unknown_object() to get the real "struct object", and pass that. Signed-off-by: Jeff King --- fsck.c | 3 ++- t/t7415-submodule-names.sh | 18 ++++++++++++++++++ 2 files changed, 20 insertions(+), 1 deletion(-) diff --git a/fsck.c b/fsck.c index bcae2c30e6..48e7e36869 100644 --- a/fsck.c +++ b/fsck.c @@ -1036,7 +1036,8 @@ int fsck_finish(struct fsck_options *options) blob = lookup_blob(oid); if (!blob) { - ret |= report(options, &blob->object, + struct object *obj = lookup_unknown_object(oid->hash); + ret |= report(options, obj, FSCK_MSG_GITMODULES_BLOB, "non-blob found at .gitmodules"); continue; diff --git a/t/t7415-submodule-names.sh b/t/t7415-submodule-names.sh index 541bd81684..3c0f1a102a 100755 --- a/t/t7415-submodule-names.sh +++ b/t/t7415-submodule-names.sh @@ -148,4 +148,22 @@ test_expect_success 'fsck detects symlinked .gitmodules file' ' ) ' +test_expect_success 'fsck detects non-blob .gitmodules' ' + git init non-blob && + ( + cd non-blob && + + # As above, make the funny tree directly to avoid index + # restrictions. + mkdir subdir && + cp ../.gitmodules subdir/file && + git add subdir/file && + git commit -m ok && + git ls-tree HEAD | sed s/subdir/.gitmodules/ | git mktree && + + test_must_fail git fsck 2>output && + grep gitmodulesBlob output + ) +' + test_done -- 2.18.0.rc1.446.g4486251e51 ``` ## Jeff King, 2018-06-12 09:51 Subject: Re: [PATCH] fsck: avoid looking at NULL blob->object Message-ID: <20180612095121.GA27181@sigill.intra.peff.net> URL: https://gitlist.dev/e/20180612095121.GA27181%40sigill.intra.peff.net In-Reply-To: ``` On Sat, Jun 09, 2018 at 03:44:30PM +0200, Martin Ågren wrote: > On 9 June 2018 at 11:21, Jeff King wrote: > > On Sat, Jun 09, 2018 at 10:50:36AM +0200, Martin Ågren wrote: > > > >> On 9 June 2018 at 10:32, Jeff King wrote: > >> > Except it _does_ do one non-trivial thing, which is call the > >> > report() function, which wants us to pass a pointer to a > >> > "struct object". Which we don't have (we have only a "struct > >> > object_id"). So we erroneously passed the NULL object, which > >> > >> s/passed/dereferenced/? Probably doesn't affect the fix though. > > > > Well, we passed it, and then that function dereferenced it. :) > > > > I'm going to re-roll for the minor bits that Eric pointed out, so I'll > > try to word this better. > > My bad. I somehow thought we get into trouble already before we call > `report()`. Well, we do, since we have undefined behavior. But for all > practical purposes `&blob->object` and `blob` are the same > (NULL-)pointer so we only crash after we call `report()`. > > Anyway, obviously no need to do anything about this in a v3. Ah, yeah, I didn't really think of it that way. But certainly you are right that the moment we look at &blob->object, we are invoking undefined behavior according to the standard. Hopefully the wording tweak I made covers both ways of thinking about it. :) -Peff ```