Re: [PATCH] dfa: don't overrun a malloc'd buffer for certain regexps

[Jim Meyering]

Stefano Lattarini wrote:
> Hi Jim.  Sorry to jump in just to be nitpicky, but ...
...
>> diff --git a/tests/Makefile.am b/tests/Makefile.am
>> index 8d51727..1f0d2cf 100644
>> --- a/tests/Makefile.am
>> +++ b/tests/Makefile.am
...
>> @@ -103,7 +104,6 @@ MALLOC_PERTURB_ = 1
>>  TESTS_ENVIRONMENT =                         \
>>    tmp__=$$TMPDIR; test -d "$$tmp__" || tmp__=.;     \
>>    TMPDIR=$$tmp__; export TMPDIR;            \
>> -  exec 9>&2;                                        \
>>    shell_or_perl_() {                                \
>>      if grep '^\#!/usr/bin/perl' "$$1" > /dev/null; then                     
>> \
>>        if $(PERL) -e 'use warnings' > /dev/null 2>&1; then           \
>> @@ -141,6 +141,6 @@ TESTS_ENVIRONMENT =                              \
>>    PERL='$(PERL)'                            \
>>    SHELL='$(SHELL)'                          \
>>    PATH='$(abs_top_builddir)/src$(PATH_SEPARATOR)'"$$PATH" \
>> -  ; shell_or_perl_
>> +  ; shell_or_perl_ 9>&2
>>
> ... wouldn't it be better to do this fix to TESTS_ENVIRONMENT in
> a follow-up patch?  It's not even mentioned in the ChangeLog.

Hi Stefano,

Thanks for the review.  Yes, that would have been better.
If I'd known someone would review carefully and promptly
I wouldn't have pushed right away.

For the record, to others who have not been following this thread
elsewhere (coreutils/gnulib/automake lists), removing the "exec 9>&2"
and instead appending 9>&2 fixes a portability nit that would cause
decreased functionality and a diagnostic when running "make check"
and using a ksh-based /bin/sh or /bin/sh on HP-UX.