From: Elijah Newren <newren@gmail.com>
Date: Wed, 29 Sep 2010 20:30:19 GMT
Subject: Re: [PATCHv4 15/15] Replace "unset VAR" with "unset VAR;" in testsuite as per t/README
Message-ID: <AANLkTimgTAerCNcSHBR9t-s-ThWHLMXdsb2T=E2w8gpB@mail.gmail.com>
In-Reply-To: <7viq1omiv8.fsf@alter.siamese.dyndns.org>

Hi,

Thanks for all the reviews.  I'll fix things up.  One question on this
last one...

On Wed, Sep 29, 2010 at 1:48 PM, Junio C Hamano <gitster@pobox.com> wrote:
> Elijah Newren <newren@gmail.com> writes:
>
>> Acked-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
>> Signed-off-by: Elijah Newren <newren@gmail.com>
>> ---
>> @@ -175,8 +175,8 @@ test_expect_success 'init with init.templatedir set' '
>>               git config -f "$test_config"  init.templatedir "${HOME}/templatedir-source" &&
>>               mkdir templatedir-set &&
>>               cd templatedir-set &&
>> -             unset GIT_CONFIG_NOGLOBAL &&
>> -             unset GIT_TEMPLATE_DIR &&
>> +             unset GIT_CONFIG_NOGLOBAL;
>> +             unset GIT_TEMPLATE_DIR;
>>               NO_SET_GIT_TEMPLATE_DIR=t &&
>>               export NO_SET_GIT_TEMPLATE_DIR &&
>>               git init
>> @@ -187,7 +187,7 @@ test_expect_success 'init with init.templatedir set' '
>>  test_expect_success 'init --bare/--shared overrides system/global config' '
>>       (
>>               test_config="$HOME"/.gitconfig &&
>> -             unset GIT_CONFIG_NOGLOBAL &&
>> +             unset GIT_CONFIG_NOGLOBAL;
>>               git config -f "$test_config" core.bare false &&
>>               git config -f "$test_config" core.sharedRepository 0640 &&
>>               mkdir init-bare-shared-override &&
>> @@ -202,7 +202,7 @@ test_expect_success 'init --bare/--shared overrides system/global config' '
>>  test_expect_success 'init honors global core.sharedRepository' '
>>       (
>>               test_config="$HOME"/.gitconfig &&
>> -             unset GIT_CONFIG_NOGLOBAL &&
>> +             unset GIT_CONFIG_NOGLOBAL;
>>               git config -f "$test_config" core.sharedRepository 0666 &&
>>               mkdir shared-honor-global &&
>>               cd shared-honor-global &&
>
> These three hunks look wrong as they break the && cascades.  All others
> (including the change to 7502) look Ok.

They break the && cascades, but aren't the && cascades already
inherently broken due to the unportable return value of unset?

> Personally I do not agree that ";" at the end is explicit enough as
> t/README seems to think.  If we want to be explicit, I'd say we should do
> so by saying something like:
>
>        unset VAR ;# can fail

That makes sense.  Alternatively, I asked in an earlier posting of
this series whether it would it make sense to add a
  portable_unset () {
    unset $* || true
  }
helper function to test-lib, and just use it instead of unset?  That
would allow proper chaining everywhere.  Thoughts?