[v2] test-lib: check Bash version for '-x' without using shell arrays
diff mbox series

Message ID 20190103114317.11523-1-szeder.dev@gmail.com
State New
Headers show
Series
  • [v2] test-lib: check Bash version for '-x' without using shell arrays
Related show

Commit Message

SZEDER Gábor Jan. 3, 2019, 11:43 a.m. UTC
One of our test scripts, 't1510-repo-setup.sh' [1], still can't be
reliably run with '-x' tracing enabled, unless it's executed with a
Bash version supporting BASH_XTRACEFD (since v4.1).  We have a lengthy
condition to check the version of the shell running the test script,
and disable tracing if it's not executed with a suitable Bash version
[2].

This condition uses non-portable shell array accesses to easily get
Bash's major and minor version number.  This didn't seem to be
problematic, because the simple commands expanding those array
accesses are only executed when the test script is actually run with
Bash.  When run with Dash, the only shell I have at hand that doesn't
support shell arrays, there are no issues, as it apparently skips
right over the non-executed simple commands without noticing the
non-supported constructs.

Alas, it has been reported that NetBSD's /bin/sh does complain about
them:

  ./test-lib.sh: 327: Syntax error: Bad substitution

where line 327 contains the first ${BASH_VERSINFO[0]} array access.

To my understanding both shells are right and conform to POSIX,
because the standard allows both behaviors by stating the following
under '2.8.1 Consequences of Shell Errors' [3]:

  "An expansion error is one that occurs when the shell expansions
  define in wordexp are carried out (for example, "${x!y}", because
  '!' is not a valid operator); an implementation may treat these as
  syntax errors if it is able to detect them during tokenization,
  rather than during expansion."

Avoid this issue with NetBSD's /bin/sh (and potentially with other,
less common shells) by hiding the shell array syntax behind 'eval'
that is only executed with Bash.

[1] 5827506928 (t1510-repo-setup: mark as untraceable with '-x',
    2018-02-24)
[2] 5fc98e79fc (t: add means to disable '-x' tracing for individual
    test scripts, 2018-02-24)
[3] http://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html#tag_18_08_01

Reported-by: Max Kirillov <max@max630.net>
Helped-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
---

Changes since v1:

 - Hide the shell array syntax behind 'eval'.
   (I'm fine with both versions, take your pick.)
 - Corrected typo in commit message that Eric pointed out.
 - Added a link to the relevant section in POSIX.

 t/test-lib.sh | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Comments

Junio C Hamano Jan. 3, 2019, 8:29 p.m. UTC | #1
SZEDER Gábor <szeder.dev@gmail.com> writes:

> One of our test scripts, 't1510-repo-setup.sh' [1], still can't be
> reliably run with '-x' tracing enabled, unless it's executed with a
> Bash version supporting BASH_XTRACEFD (since v4.1).  We have a lengthy
> condition to check the version of the shell running the test script,
> and disable tracing if it's not executed with a suitable Bash version
> [2].
>
> This condition uses non-portable shell array accesses to easily get
> Bash's major and minor version number.  This didn't seem to be
> problematic, because the simple commands expanding those array
> accesses are only executed when the test script is actually run with
> Bash.  When run with Dash, the only shell I have at hand that doesn't
> support shell arrays, there are no issues, as it apparently skips
> right over the non-executed simple commands without noticing the
> non-supported constructs.
>
> Alas, it has been reported that NetBSD's /bin/sh does complain about
> them:
>
>   ./test-lib.sh: 327: Syntax error: Bad substitution
>
> where line 327 contains the first ${BASH_VERSINFO[0]} array access.
>
> To my understanding both shells are right and conform to POSIX,
> because the standard allows both behaviors by stating the following
> under '2.8.1 Consequences of Shell Errors' [3]:
>
>   "An expansion error is one that occurs when the shell expansions
>   define in wordexp are carried out (for example, "${x!y}", because
>   '!' is not a valid operator); an implementation may treat these as
>   syntax errors if it is able to detect them during tokenization,
>   rather than during expansion."
>
> Avoid this issue with NetBSD's /bin/sh (and potentially with other,
> less common shells) by hiding the shell array syntax behind 'eval'
> that is only executed with Bash.
>
> [1] 5827506928 (t1510-repo-setup: mark as untraceable with '-x',
>     2018-02-24)
> [2] 5fc98e79fc (t: add means to disable '-x' tracing for individual
>     test scripts, 2018-02-24)
> [3] http://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html#tag_18_08_01
>
> Reported-by: Max Kirillov <max@max630.net>
> Helped-by: Johannes Sixt <j6t@kdbg.org>
> Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
> ---
>
> Changes since v1:
>
>  - Hide the shell array syntax behind 'eval'.
>    (I'm fine with both versions, take your pick.)
>  - Corrected typo in commit message that Eric pointed out.
>  - Added a link to the relevant section in POSIX.

Let's treat this as an independent and more urgent fix-up.  I think
it is sufficient to apply it to 2.20.x track, even though we could
go back to 2.17.x and above.

And then let's tentatively kick the "stress test" series out of
'pu', and have that series rebuilt on top of 'master' and this
patch.

Thanks.

>
>  t/test-lib.sh | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/t/test-lib.sh b/t/test-lib.sh
> index 0f1faa24b2..c34831a4de 100644
> --- a/t/test-lib.sh
> +++ b/t/test-lib.sh
> @@ -323,12 +323,12 @@ do
>  		# this test is marked as such, and ignore '-x' if it
>  		# isn't executed with a suitable Bash version.
>  		if test -z "$test_untraceable" || {
> -		     test -n "$BASH_VERSION" && {
> +		     test -n "$BASH_VERSION" && eval '
>  		       test ${BASH_VERSINFO[0]} -gt 4 || {
>  			 test ${BASH_VERSINFO[0]} -eq 4 &&
>  			 test ${BASH_VERSINFO[1]} -ge 1
>  		       }
> -		     }
> +		     '
>  		   }
>  		then
>  			trace=t
SZEDER Gábor Jan. 4, 2019, 9:30 a.m. UTC | #2
On Thu, Jan 03, 2019 at 12:29:35PM -0800, Junio C Hamano wrote:
> Let's treat this as an independent and more urgent fix-up.  I think
> it is sufficient to apply it to 2.20.x track, even though we could
> go back to 2.17.x and above.
> 
> And then let's tentatively kick the "stress test" series out of
> 'pu', and have that series rebuilt on top of 'master' and this
> patch.

I rebased my '--stress' patch series on top of
'sg/test-bash-version-fix', and the result is the same as what's at
the tip of 'sg/stress-test' at 1d1416a34b.
Junio C Hamano Jan. 4, 2019, 11:25 a.m. UTC | #3
SZEDER Gábor <szeder.dev@gmail.com> writes:

> On Thu, Jan 03, 2019 at 12:29:35PM -0800, Junio C Hamano wrote:
>> Let's treat this as an independent and more urgent fix-up.  I think
>> it is sufficient to apply it to 2.20.x track, even though we could
>> go back to 2.17.x and above.
>> 
>> And then let's tentatively kick the "stress test" series out of
>> 'pu', and have that series rebuilt on top of 'master' and this
>> patch.
>
> I rebased my '--stress' patch series on top of
> 'sg/test-bash-version-fix', and the result is the same as what's at
> the tip of 'sg/stress-test' at 1d1416a34b.

Yeah, sorry for not reporting what I did after pushing the
integration result out.  I ended up rebasing the stress series on
top of the updated fix with eval myself.

Thanks.
SZEDER Gábor Jan. 4, 2019, 12:38 p.m. UTC | #4
On Fri, Jan 04, 2019 at 03:25:57AM -0800, Junio C Hamano wrote:
> SZEDER Gábor <szeder.dev@gmail.com> writes:
> 
> > On Thu, Jan 03, 2019 at 12:29:35PM -0800, Junio C Hamano wrote:
> >> Let's treat this as an independent and more urgent fix-up.  I think
> >> it is sufficient to apply it to 2.20.x track, even though we could
> >> go back to 2.17.x and above.
> >> 
> >> And then let's tentatively kick the "stress test" series out of
> >> 'pu', and have that series rebuilt on top of 'master' and this
> >> patch.
> >
> > I rebased my '--stress' patch series on top of
> > 'sg/test-bash-version-fix', and the result is the same as what's at
> > the tip of 'sg/stress-test' at 1d1416a34b.
> 
> Yeah, sorry for not reporting what I did after pushing the
> integration result out.  I ended up rebasing the stress series on
> top of the updated fix with eval myself.

No problem, it still saved me from writing a cover letter :)

Note that I didn't actually test either of these patches on NetBSD,
though both "Should Work".  Carlo sent a Tested-by for v1; I hope
he'll be able to get around and test v2 as well.
Carlo Marcelo Arenas Belón Jan. 4, 2019, 6:42 p.m. UTC | #5
v2 works fine, as expected

Carlo
Junio C Hamano Jan. 4, 2019, 8:04 p.m. UTC | #6
Carlo Arenas <carenas@gmail.com> writes:

> v2 works fine, as expected

Thanks.

Patch
diff mbox series

diff --git a/t/test-lib.sh b/t/test-lib.sh
index 0f1faa24b2..c34831a4de 100644
--- a/t/test-lib.sh
+++ b/t/test-lib.sh
@@ -323,12 +323,12 @@  do
 		# this test is marked as such, and ignore '-x' if it
 		# isn't executed with a suitable Bash version.
 		if test -z "$test_untraceable" || {
-		     test -n "$BASH_VERSION" && {
+		     test -n "$BASH_VERSION" && eval '
 		       test ${BASH_VERSINFO[0]} -gt 4 || {
 			 test ${BASH_VERSINFO[0]} -eq 4 &&
 			 test ${BASH_VERSINFO[1]} -ge 1
 		       }
-		     }
+		     '
 		   }
 		then
 			trace=t