Re: [PATCH 1/2] stdio-common: run AWK in the C locale in the printf format tests

Adhemerval Zanella Netto <[email protected]> Tue, 4 Aug 2026 10:37:28 -0300
Newsgroups gmane.comp.lib.glibc.alpha
Organization Linaro
Message-ID <[email protected]>

On 04/08/26 10:18, Adhemerval Zanella Netto wrote:
> 
> 
> On 03/08/26 22:59, Matt Turner wrote:
>> The program under test runs in the C locale, through the test program
>> prefix, but AWK inherits whatever locale the build was started in.  They
>> agree today only because the locale in use shares its decimal point with
>> the C locale.
>>
>> It is also faster.  gawk takes a single byte path in its regular
>> expression engine when MB_CUR_MAX is 1, and the script matches several
>> expressions against every line.  For the %f conversion for double, the
>> largest of these tests, as the median of five runs:
>>
>>   x86_64, gawk 5.4.1    1.482s -> 1.248s
>>   x86_64, gawk 5.3.2    0.911s -> 0.703s
>>   alpha,  gawk 5.4.60    30.9s ->  26.6s
>>
>> Worth noting that gawk 5.4 is a good deal slower here than 5.3 was, at
>> 1.248s against 0.703s for the same input in the C locale, so these tests
>> have become more expensive than they used to be.
> 
> LGTM, thanks.
> 
> Reviewed-by: Adhemerval Zanella  <[email protected]>

Scratch that, testing shows that LC_ALL=C expansion is not working correctly:

$ cat stdio-common/tst-printf-format-as-c.out
Verifying c
Conversion c output error, first line:
tst-printf-format-c.sh: line 32: LC_ALL=C: command not found

The scripts invoke '$AWK -f tst-printf-format.awk', shell recognizes VAR=value 
assignment prefixes at parse time, before variable expansion. You need to use
'env' here:

  AWK="env LC_ALL=C ${AWK:-awk}"

> 
>> ---
>>  stdio-common/tst-printf-format-c.sh       | 2 +-
>>  stdio-common/tst-printf-format-char.sh    | 2 +-
>>  stdio-common/tst-printf-format-double.sh  | 2 +-
>>  stdio-common/tst-printf-format-int.sh     | 2 +-
>>  stdio-common/tst-printf-format-ldouble.sh | 2 +-
>>  stdio-common/tst-printf-format-llong.sh   | 2 +-
>>  stdio-common/tst-printf-format-long.sh    | 2 +-
>>  stdio-common/tst-printf-format-s.sh       | 2 +-
>>  stdio-common/tst-printf-format-short.sh   | 2 +-
>>  stdio-common/tst-printf-format-uchar.sh   | 2 +-
>>  stdio-common/tst-printf-format-uint.sh    | 2 +-
>>  stdio-common/tst-printf-format-ullong.sh  | 2 +-
>>  stdio-common/tst-printf-format-ulong.sh   | 2 +-
>>  stdio-common/tst-printf-format-ushort.sh  | 2 +-
>>  14 files changed, 14 insertions(+), 14 deletions(-)
>>
>> diff --git ./stdio-common/tst-printf-format-c.sh ./stdio-common/tst-printf-format-c.sh
>> index 825c50ec42..636884d3a7 100644
>> --- ./stdio-common/tst-printf-format-c.sh
>> +++ ./stdio-common/tst-printf-format-c.sh
>> @@ -23,7 +23,7 @@ xprintf=$1; shift
>>  common_objpfx=$1; shift
>>  test_program_prefix=$1; shift
>>  
>> -AWK=${AWK:-awk}
>> +AWK="LC_ALL=C ${AWK:-awk}"
>>  
>>  echo Verifying c
>>  (set -o pipefail
>> diff --git ./stdio-common/tst-printf-format-char.sh ./stdio-common/tst-printf-format-char.sh
>> index 7867bdd62f..a3eb0665f2 100644
>> --- ./stdio-common/tst-printf-format-char.sh
>> +++ ./stdio-common/tst-printf-format-char.sh
>> @@ -23,7 +23,7 @@ xprintf=$1; shift
>>  common_objpfx=$1; shift
>>  test_program_prefix=$1; shift
>>  
>> -AWK=${AWK:-awk}
>> +AWK="LC_ALL=C ${AWK:-awk}"
>>  
>>  status=0
>>  
>> diff --git ./stdio-common/tst-printf-format-double.sh ./stdio-common/tst-printf-format-double.sh
>> index 8157092dc4..ddbffd19aa 100644
>> --- ./stdio-common/tst-printf-format-double.sh
>> +++ ./stdio-common/tst-printf-format-double.sh
>> @@ -29,7 +29,7 @@ test_program_prefix=$1; shift
>>  # internally to process the conversion requested, so any bug in our code
>>  # would then be verified against itself, defeating the objective of doing
>>  # the verification against an independent implementation.
>> -AWK="${AWK:-awk} -M"
>> +AWK="LC_ALL=C ${AWK:-awk} -M"
>>  
>>  status=77
>>  
>> diff --git ./stdio-common/tst-printf-format-int.sh ./stdio-common/tst-printf-format-int.sh
>> index 8542ff4150..2cca1dc884 100644
>> --- ./stdio-common/tst-printf-format-int.sh
>> +++ ./stdio-common/tst-printf-format-int.sh
>> @@ -23,7 +23,7 @@ xprintf=$1; shift
>>  common_objpfx=$1; shift
>>  test_program_prefix=$1; shift
>>  
>> -AWK=${AWK:-awk}
>> +AWK="LC_ALL=C ${AWK:-awk}"
>>  
>>  status=77
>>  
>> diff --git ./stdio-common/tst-printf-format-ldouble.sh ./stdio-common/tst-printf-format-ldouble.sh
>> index dbf78a98c6..9acb1237de 100644
>> --- ./stdio-common/tst-printf-format-ldouble.sh
>> +++ ./stdio-common/tst-printf-format-ldouble.sh
>> @@ -29,7 +29,7 @@ test_program_prefix=$1; shift
>>  # internally to process the conversion requested, so any bug in our code
>>  # would then be verified against itself, defeating the objective of doing
>>  # the verification against an independent implementation.
>> -AWK="${AWK:-awk} -M"
>> +AWK="LC_ALL=C ${AWK:-awk} -M"
>>  
>>  status=77
>>  
>> diff --git ./stdio-common/tst-printf-format-llong.sh ./stdio-common/tst-printf-format-llong.sh
>> index e1f5252c9a..5e10e87fa6 100644
>> --- ./stdio-common/tst-printf-format-llong.sh
>> +++ ./stdio-common/tst-printf-format-llong.sh
>> @@ -23,7 +23,7 @@ xprintf=$1; shift
>>  common_objpfx=$1; shift
>>  test_program_prefix=$1; shift
>>  
>> -AWK=${AWK:-awk}
>> +AWK="LC_ALL=C ${AWK:-awk}"
>>  
>>  status=77
>>  
>> diff --git ./stdio-common/tst-printf-format-long.sh ./stdio-common/tst-printf-format-long.sh
>> index 4b68ab1e04..0e17fcd596 100644
>> --- ./stdio-common/tst-printf-format-long.sh
>> +++ ./stdio-common/tst-printf-format-long.sh
>> @@ -23,7 +23,7 @@ xprintf=$1; shift
>>  common_objpfx=$1; shift
>>  test_program_prefix=$1; shift
>>  
>> -AWK=${AWK:-awk}
>> +AWK="LC_ALL=C ${AWK:-awk}"
>>  
>>  status=77
>>  
>> diff --git ./stdio-common/tst-printf-format-s.sh ./stdio-common/tst-printf-format-s.sh
>> index 65aa0cb675..d0a2ffd166 100644
>> --- ./stdio-common/tst-printf-format-s.sh
>> +++ ./stdio-common/tst-printf-format-s.sh
>> @@ -23,7 +23,7 @@ xprintf=$1; shift
>>  common_objpfx=$1; shift
>>  test_program_prefix=$1; shift
>>  
>> -AWK=${AWK:-awk}
>> +AWK="LC_ALL=C ${AWK:-awk}"
>>  
>>  echo Verifying s
>>  (set -o pipefail
>> diff --git ./stdio-common/tst-printf-format-short.sh ./stdio-common/tst-printf-format-short.sh
>> index 30357baa02..bd759b5bb7 100644
>> --- ./stdio-common/tst-printf-format-short.sh
>> +++ ./stdio-common/tst-printf-format-short.sh
>> @@ -23,7 +23,7 @@ xprintf=$1; shift
>>  common_objpfx=$1; shift
>>  test_program_prefix=$1; shift
>>  
>> -AWK=${AWK:-awk}
>> +AWK="LC_ALL=C ${AWK:-awk}"
>>  
>>  status=0
>>  
>> diff --git ./stdio-common/tst-printf-format-uchar.sh ./stdio-common/tst-printf-format-uchar.sh
>> index 08a6914b88..6f41009ad3 100644
>> --- ./stdio-common/tst-printf-format-uchar.sh
>> +++ ./stdio-common/tst-printf-format-uchar.sh
>> @@ -23,7 +23,7 @@ xprintf=$1; shift
>>  common_objpfx=$1; shift
>>  test_program_prefix=$1; shift
>>  
>> -AWK=${AWK:-awk}
>> +AWK="LC_ALL=C ${AWK:-awk}"
>>  
>>  status=0
>>  
>> diff --git ./stdio-common/tst-printf-format-uint.sh ./stdio-common/tst-printf-format-uint.sh
>> index 0ba203ccda..be2da0064b 100644
>> --- ./stdio-common/tst-printf-format-uint.sh
>> +++ ./stdio-common/tst-printf-format-uint.sh
>> @@ -23,7 +23,7 @@ xprintf=$1; shift
>>  common_objpfx=$1; shift
>>  test_program_prefix=$1; shift
>>  
>> -AWK=${AWK:-awk}
>> +AWK="LC_ALL=C ${AWK:-awk}"
>>  
>>  status=77
>>  
>> diff --git ./stdio-common/tst-printf-format-ullong.sh ./stdio-common/tst-printf-format-ullong.sh
>> index 5b881ab924..45dfafb682 100644
>> --- ./stdio-common/tst-printf-format-ullong.sh
>> +++ ./stdio-common/tst-printf-format-ullong.sh
>> @@ -23,7 +23,7 @@ xprintf=$1; shift
>>  common_objpfx=$1; shift
>>  test_program_prefix=$1; shift
>>  
>> -AWK=${AWK:-awk}
>> +AWK="LC_ALL=C ${AWK:-awk}"
>>  
>>  status=77
>>  
>> diff --git ./stdio-common/tst-printf-format-ulong.sh ./stdio-common/tst-printf-format-ulong.sh
>> index f6aeb8e3c0..24e36cbccf 100644
>> --- ./stdio-common/tst-printf-format-ulong.sh
>> +++ ./stdio-common/tst-printf-format-ulong.sh
>> @@ -23,7 +23,7 @@ xprintf=$1; shift
>>  common_objpfx=$1; shift
>>  test_program_prefix=$1; shift
>>  
>> -AWK=${AWK:-awk}
>> +AWK="LC_ALL=C ${AWK:-awk}"
>>  
>>  status=77
>>  
>> diff --git ./stdio-common/tst-printf-format-ushort.sh ./stdio-common/tst-printf-format-ushort.sh
>> index 07609128ab..0e1008c75e 100644
>> --- ./stdio-common/tst-printf-format-ushort.sh
>> +++ ./stdio-common/tst-printf-format-ushort.sh
>> @@ -23,7 +23,7 @@ xprintf=$1; shift
>>  common_objpfx=$1; shift
>>  test_program_prefix=$1; shift
>>  
>> -AWK=${AWK:-awk}
>> +AWK="LC_ALL=C ${AWK:-awk}"
>>  
>>  status=0
>>  
>