Re: [PATCH] init.sh: don't waste a subshell just to redirect stderr

[Jim Meyering]

Stefano Lattarini wrote:
> Hi Jim.
>
> On 01/05/2012 04:46 PM, Jim Meyering wrote:
>> Looking at Bruno's recent test-init.sh failure,
>> I noticed the unnecessary sub-shells here:
>>
>> From a815c2160e6d6cc27d934e669ef9486bab310db3 Mon Sep 17 00:00:00 2001
>> From: Jim Meyering <address@hidden>
>> Date: Thu, 5 Jan 2012 16:45:50 +0100
>> Subject: [PATCH] init.sh: don't waste a subshell just to redirect stderr
>>
>> * tests/init.sh: In testing for diff -u and diff -c, use a
>> stderr-redirecting exec inside `...` rather than a subshell.
>> ---
>>  ChangeLog     |    4 ++++
>>  tests/init.sh |    4 ++--
>>  2 files changed, 6 insertions(+), 2 deletions(-)
>>
>> diff --git a/ChangeLog b/ChangeLog
>> index eb4a3a1..3eb41be 100644
>> --- a/ChangeLog
>> +++ b/ChangeLog
>> @@ -1,5 +1,9 @@
>>  2012-01-05  Jim Meyering  <address@hidden>
>>
>> +    init.sh: don't waste a subshell just to redirect stderr
>> +    * tests/init.sh: In testing for diff -u and diff -c, use a
>> +    stderr-redirecting exec inside `...` rather than a subshell.
>> +
>>      test-init.sh: avoid failure on HP-UX 11.00
>>      * tests/test-init.sh: Skip "diff -u"-comparing step when compare
>>      resolves to diff -c or cmp.  Reported by Bruno Haible.
>> diff --git a/tests/init.sh b/tests/init.sh
>> index 9485492..65458c7 100644
>> --- a/tests/init.sh
>> +++ b/tests/init.sh
>> @@ -255,7 +255,7 @@ compare_dev_null_ ()
>>    return 2
>>  }
>>
>> -if diff_out_=`( diff -u "$0" "$0" < /dev/null ) 2>/dev/null`; then
>> +if diff_out_=`exec 2>/dev/null; diff -u "$0" "$0" < /dev/null`; then
>>
> Wouldn't it be simpler and better to leave that redirection out altogether?
> The output of the test scripts should be redirected to `.log' files by the
> testsuite harness, so by being more verbose we can make those files more
> informative, all without cluttering the make output.

It depends.
In many cases, I would agree with you, but probably not here.

I think it comes down to balancing the small repetitive cost of an
added diagnostic in every log on systems lacking diff -u support vs.
the probability that the clutter will ever be useful.