Re: Bison 3.5.93 on AIX 7.2

[Akim Demaille]

Hi Bruno!

> Le 3 mai 2020 à 22:35, Bruno Haible <address@hidden> a écrit :
> 
> On AIX 7.2, when building with gcc in 32-bit mode, there are three problems:
> 
> 1) FAIL: examples/c/reccalc/reccalc.test
> As reported in
> <https://lists.gnu.org/archive/html/bug-bison/2020-05/msg00047.html> .

This is the seq issue, which is hopefully now addressed by 
https://lists.gnu.org/archive/html/bug-bison/2020-05/msg00060.html.

> 2) "gmake -k check" goes past 1) and starts to execute the ca. 600 individual
> tests, but it aborts like this:
> 
> 276: Several parsers                                 ok
> ../tests/testsuite[25]: syntax error at line 3 : `)' unexpected
> Makefile:11650: recipe for target 'check-local' failed
> gmake[3]: *** [check-local] Error 2
> 
> The workaround is to change, in the Makefile,
>  SHELL = /bin/sh
> to
>  SHELL = /bin/bash
> 
> Then "gmake -k check" executes the full test suite.
> 
> I added a 'set -x' at the top of tests/testsuite, ran "gmake -k check" with
> and without the change, and compared the two outputs. But it gives no clue
> why /bin/sh chokes in the middle. 

Bummer...

Where you running the test suite in sequential mode?  Was 276 really the last 
successful test, and then it crashed?


> 3) There are three test failures:
> 137: Tabulations and multibyte characters            FAILED 
> (diagnostics.at:280)
> 527: GNU AWK 3.1.0 Grammar: IELR(1)                  FAILED (existing.at:801)
> 533: GNU pic (Groff 1.18.1) Grammar: IELR(1)         FAILED (existing.at:3266)
> 
> Detailed configuration:
> 
> CC="gcc"; CXX="g++"; export CC CXX
> unset AR NM
> 
> Detailed test results: attached.
> 
> The first test failure is like on Cygwin: because wchar_t is only 16 bits
> wide, in 32-bit AIX.

Ok, so same "cure".  I don't know how to check if we are in 32-bit AIX, but I 
don't care much about running this test everywhere (the "safe" part is still 
run as test 136), so let's just skip on AIX.

commit 0fafbbdefb288cda73990a156cf185e8fd0cf04b
Author: Akim Demaille <address@hidden>
Date:   Mon May 4 08:07:48 2020 +0200

    tests: beware of wchar_t portability issues on AIX
    
    https://lists.gnu.org/r/bug-bison/2020-05/msg00050.html
    Reported by Bruno Haible.
    
    * tests/diagnostics.at: here.

diff --git a/tests/diagnostics.at b/tests/diagnostics.at
index 46af89c4..035948c4 100644
--- a/tests/diagnostics.at
+++ b/tests/diagnostics.at
@@ -265,7 +265,8 @@ input.y: <warning>warning:</warning> fix-its can be 
applied.  Rerun with option
 ]])
 
 
-# Likewise, but currently not portable to Cygwin.
+# Likewise, but currently not portable to AIX and Cygwin.
+# https://lists.gnu.org/r/bug-bison/2020-05/msg00050.html
 # https://lists.gnu.org/r/bug-bison/2020-05/msg00003.html.
 AT_TEST([[Tabulations and multibyte characters]],
 [[%%
@@ -277,7 +278,7 @@ e: {∇⃗×𝐸⃗ = -∂𝐵⃗/∂t}
       |    <warning>^~~~~~~~~~~~~~</warning>
       |    <fixit-insert>%empty</fixit-insert>
 input.y: <warning>warning:</warning> fix-its can be applied.  Rerun with 
option '--update'. [<warning>-Wother</warning>]
-]], [], [uname -a | grep -i cygwin])
+]], [], [uname -a | $EGREP -i 'aix|cygwin'])
 
 
 ## --------------- ##


> The two other test failures are like on AIX 7.1, see
> <https://lists.gnu.org/archive/html/bug-bison/2020-05/msg00049.html>.

That is to say, the "diff -u" portability problems, addressed in 
https://lists.gnu.org/r/bug-bison/2020-05/msg00059.html.

Thanks Bruno!