threads / patch / 49125

patchsubmodule: add more exhaustive up-path testing

Subject: [PATCH] submodule: add more exhaustive up-path testing

## tl;dr

7 messages between Aug 14, 2018 and Aug 14, 2018. Diffs are folded; open one to read it.

replies: 6people: 3as markdown or json

Ævar Arnfjörð Bjarmason· Aug 14, 2018, 18:59 UTC · lore

The tests added in 63e95beb08 ("submodule: port resolve_relative_url from shell to C", 2016-04-15) didn't do a good job of testing various up-path invocations where the up-path would bring us beyond even the URL in question without emitting an error.

These results look nonsensical, but it's worth exhaustively testing them before fixing any of this code, so we can see which of these cases were changed.

Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
---

The reason I started to poke at this was because I wanted to add support for relative URLs like //group/project.

This solves the problem of there being e.g. a myhost.com/group/project which includes a ../submodule submodule pointing to myhost.com/group/submodule , and me wanting to fork it to myhost.com/myuser/project.

But of course with ../submodule the relative URL will point to myhost.com/myuser/submodule, which won't exist.

So I was hoping to find some point in the relative URL code where we were actually aware of what the host boundary was, so I could just cut off everything after that if //group/project was provided, strip off one slash to make it /group/project, and call it a day.

But as the tests below show the code isn't aware of that at all, and will happily trample over the host(name) to produce nonsensical results.

So I think these tests are worthwihle in themselves, but would like some advice on how to proceed with that from someone more familiar with submodules.

 t/t0060-path-utils.sh | 13 +++++++++++++
 1 file changed, 13 insertions(+)
Show changes to t/t0060-path-utils.sh +13 −0
diff --git a/t/t0060-path-utils.sh b/t/t0060-path-utils.sh
index 21a8b53132..cd74c0a471 100755
--- a/t/t0060-path-utils.sh
+++ b/t/t0060-path-utils.sh
@@ -330,6 +330,9 @@ test_submodule_relative_url "(null)" "../foo" "../submodule" "../submodule"
 test_submodule_relative_url "(null)" "./foo/bar" "../submodule" "foo/submodule"
 test_submodule_relative_url "(null)" "./foo" "../submodule" "submodule"
 test_submodule_relative_url "(null)" "//somewhere else/repo" "../subrepo" "//somewhere else/subrepo"
+test_submodule_relative_url "(null)" "//somewhere else/repo" "../../subrepo" "//subrepo"
+test_submodule_relative_url "(null)" "//somewhere else/repo" "../../../subrepo" "/subrepo"
+test_submodule_relative_url "(null)" "//somewhere else/repo" "../../../../subrepo" "subrepo"
 test_submodule_relative_url "(null)" "$(pwd)/subsuper_update_r" "../subsubsuper_update_r" "$(pwd)/subsubsuper_update_r"
 test_submodule_relative_url "(null)" "$(pwd)/super_update_r2" "../subsuper_update_r" "$(pwd)/subsuper_update_r"
 test_submodule_relative_url "(null)" "$(pwd)/." "../." "$(pwd)/."
@@ -344,10 +347,20 @@ test_submodule_relative_url "(null)" "file:///tmp/repo" "../subrepo" "file:///tm
 test_submodule_relative_url "(null)" "foo/bar" "../submodule" "foo/submodule"
 test_submodule_relative_url "(null)" "foo" "../submodule" "submodule"
 test_submodule_relative_url "(null)" "helper:://hostname/repo" "../subrepo" "helper:://hostname/subrepo"
+test_submodule_relative_url "(null)" "helper:://hostname/repo" "../../subrepo" "helper:://subrepo"
+test_submodule_relative_url "(null)" "helper:://hostname/repo" "../../../subrepo" "helper::/subrepo"
+test_submodule_relative_url "(null)" "helper:://hostname/repo" "../../../../subrepo" "helper::subrepo"
+test_submodule_relative_url "(null)" "helper:://hostname/repo" "../../../../../subrepo" "helper:subrepo"
+test_submodule_relative_url "(null)" "helper:://hostname/repo" "../../../../../../subrepo" ".:subrepo"
 test_submodule_relative_url "(null)" "ssh://hostname/repo" "../subrepo" "ssh://hostname/subrepo"
+test_submodule_relative_url "(null)" "ssh://hostname/repo" "../../subrepo" "ssh://subrepo"
+test_submodule_relative_url "(null)" "ssh://hostname/repo" "../../../subrepo" "ssh:/subrepo"
+test_submodule_relative_url "(null)" "ssh://hostname/repo" "../../../../subrepo" "ssh:subrepo"
+test_submodule_relative_url "(null)" "ssh://hostname/repo" "../../../../../subrepo" ".:subrepo"
 test_submodule_relative_url "(null)" "ssh://hostname:22/repo" "../subrepo" "ssh://hostname:22/subrepo"
 test_submodule_relative_url "(null)" "user@host:path/to/repo" "../subrepo" "user@host:path/to/subrepo"
 test_submodule_relative_url "(null)" "user@host:repo" "../subrepo" "user@host:subrepo"
+test_submodule_relative_url "(null)" "user@host:repo" "../../subrepo" ".:subrepo"
 
 test_expect_success 'match .gitmodules' '
 	test-tool path-utils is_dotgitmodules \

-- 
2.18.0.865.gffc8e1a3cd6
Stefan Beller· Aug 14, 2018, 19:10 UTC · re: Ævar Arnfjörð Bjarmason · lore

Re: [PATCH] submodule: add more exhaustive up-path testing

On Tue, Aug 14, 2018 at 11:59 AM Ævar Arnfjörð Bjarmason <avarab@gmail.com> wrote:

Show 9 quoted lines
>
> The tests added in 63e95beb08 ("submodule: port resolve_relative_url
> from shell to C", 2016-04-15) didn't do a good job of testing various
> up-path invocations where the up-path would bring us beyond even the
> URL in question without emitting an error.
>
> These results look nonsensical, but it's worth exhaustively testing
> them before fixing any of this code, so we can see which of these
> cases were changed.

Yeah. Please look at the comment in builtin/submodule--helper.c in that commit, where I described my expectations.

I should have put them into tests instead with the expectations spelled out there.

Thanks for this patch! Stefan

>
> Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
> ---
> So I think these tests are worthwihle in themselves,

The reason I put it in the comment instead of tests was the ease of spelling out both the status quo and expectations.

> but would like
> some advice on how to proceed with that from someone more familiar
> with submodules.
So ideally we'd also error out as soon as the host name is touched?
Junio C Hamano· Aug 14, 2018, 19:54 UTC · re: Stefan Beller · lore

Re: [PATCH] submodule: add more exhaustive up-path testing

Stefan Beller <sbeller@google.com> writes:
> Thanks for this patch!
> Stefan

Thanks, I'd take it as your Acked-by: (please holler if it isn't before the patch hits 'next').

Ævar Arnfjörð Bjarmason· Aug 14, 2018, 21:05 UTC · re: Stefan Beller · lore

Re: [PATCH] submodule: add more exhaustive up-path testing

On Tue, Aug 14 2018, Stefan Beller wrote:
Show 17 quoted lines
> On Tue, Aug 14, 2018 at 11:59 AM Ævar Arnfjörð Bjarmason
> <avarab@gmail.com> wrote:
>>
>> The tests added in 63e95beb08 ("submodule: port resolve_relative_url
>> from shell to C", 2016-04-15) didn't do a good job of testing various
>> up-path invocations where the up-path would bring us beyond even the
>> URL in question without emitting an error.
>>
>> These results look nonsensical, but it's worth exhaustively testing
>> them before fixing any of this code, so we can see which of these
>> cases were changed.
>
> Yeah. Please look at the comment in builtin/submodule--helper.c
> in that commit, where I described my expectations.
>
> I should have put them into tests instead with the expectations
> spelled out there.

I'll check that out thanks. I saw that comment, but have been skimming most of this code...

Show 18 quoted lines
> Thanks for this patch!
> Stefan
>
>>
>> Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
>> ---
>
>
>> So I think these tests are worthwihle in themselves,
>
> The reason I put it in the comment instead of tests was the
> ease of spelling out both the status quo and expectations.
>
>> but would like
>> some advice on how to proceed with that from someone more familiar
>> with submodules.
>
> So ideally we'd also error out as soon as the host name is touched?

Do we have some utility function that'll take whatever we have in remote.<name>.url and spew out the username / host / path? We must, since the clone protocol needs it, but I haven't found it.

Stefan Beller· Aug 14, 2018, 21:10 UTC · re: Ævar Arnfjörð Bjarmason · lore

Re: [PATCH] submodule: add more exhaustive up-path testing

On Tue, Aug 14, 2018 at 2:05 PM Ævar Arnfjörð Bjarmason <avarab@gmail.com> wrote:

Show 5 quoted lines
> > So ideally we'd also error out as soon as the host name is touched?
>
> Do we have some utility function that'll take whatever we have in
> remote.<name>.url and spew out the username / host / path? We must,
> since the clone protocol needs it, but I haven't found it.

Nope. Welcome to the wonderful world of submodules. As submodules are in the transition state towards "in C", we could do some refactorings that would allow for easy access to these properties, but the transition is done via faithful conversion from shell to C, which did not have these functions either, but could just do some string hackery. And that is how we ended up with this function in the first place.

Ævar Arnfjörð Bjarmason· Aug 14, 2018, 21:16 UTC · re: Stefan Beller · lore

Re: [PATCH] submodule: add more exhaustive up-path testing

On Tue, Aug 14 2018, Stefan Beller wrote:
Show 16 quoted lines
> On Tue, Aug 14, 2018 at 2:05 PM Ævar Arnfjörð Bjarmason
> <avarab@gmail.com> wrote:
>
>> > So ideally we'd also error out as soon as the host name is touched?
>>
>> Do we have some utility function that'll take whatever we have in
>> remote.<name>.url and spew out the username / host / path? We must,
>> since the clone protocol needs it, but I haven't found it.
>
> Nope. Welcome to the wonderful world of submodules.
> As submodules are in the transition state towards "in C",
> we could do some refactorings that would allow for easy access to
> these properties, but the transition is done via faithful conversion from
> shell to C, which did not have these functions either, but could just
> do some string hackery. And that is how we ended up with this
> function in the first place.

The remote/clone machinery is in C and so is the resolve-relative-url helper. What's the missing bridge here?

Maybe we don't have that function yet, but we can presumably expose it, and this seems unrelated to some other bits of submodules being in shellscript.

No?
Stefan Beller· Aug 14, 2018, 21:32 UTC · re: Ævar Arnfjörð Bjarmason · lore

Re: [PATCH] submodule: add more exhaustive up-path testing

On Tue, Aug 14, 2018 at 2:16 PM Ævar Arnfjörð Bjarmason <avarab@gmail.com> wrote:

Show 29 quoted lines
>
>
> On Tue, Aug 14 2018, Stefan Beller wrote:
>
> > On Tue, Aug 14, 2018 at 2:05 PM Ævar Arnfjörð Bjarmason
> > <avarab@gmail.com> wrote:
> >
> >> > So ideally we'd also error out as soon as the host name is touched?
> >>
> >> Do we have some utility function that'll take whatever we have in
> >> remote.<name>.url and spew out the username / host / path? We must,
> >> since the clone protocol needs it, but I haven't found it.
> >
> > Nope. Welcome to the wonderful world of submodules.
> > As submodules are in the transition state towards "in C",
> > we could do some refactorings that would allow for easy access to
> > these properties, but the transition is done via faithful conversion from
> > shell to C, which did not have these functions either, but could just
> > do some string hackery. And that is how we ended up with this
> > function in the first place.
>
> The remote/clone machinery is in C and so is the resolve-relative-url
> helper. What's the missing bridge here?
>
> Maybe we don't have that function yet, but we can presumably expose it,
> and this seems unrelated to some other bits of submodules being in
> shellscript.
>
> No?
Gah, I misread your original question. Sorry about that.

← back to recent threads