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 10:18:26 -0300
Newsgroups gmane.comp.lib.glibc.alpha
Organization Linaro
Message-ID <[email protected]>

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]>

> ---
>  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
>