[PATCH 0/2] Improve printf-corners coverage

[Maciej W. Rozycki]

Hi,

 Here's a small patch series to improve printf-corners test case coverage.  

 Specifically 1/2 causes the test case to be run possibly twice, without 
and, where supported, with -M option.  This is because we have different 
code paths used for non-bignum and bignum invocations respectively, and 
then actually unexpected discrepancies in output between the two.  Both
code paths obviously need to be covered then.

 Then 2/2 adds further cases, some of which we don't currently handle 
correctly.

 I can see GAWK uses a very terse style of commits without giving them any
change descriptions, so I have followed it accordingly, giving the patches 
change headings only.

 The individual changes from the patch series can be readily applied by 
feeding the respective e-mail messages to `git am'; please respect the 
correct author attribution if applying via other means.  Changes to 
ChangeLog files are bound to cause a merge conflict, so I've included 
suitable ChangeLog entries in the respective change discussions instead, 
to be pasted by hand.

 Both changes report new failures, so GAWK has to be fixed first before 
they go in.  You can see with the series applied how initial failures are 
reported for different lines between non-bignum and bignum invocations:

printf-corners
.../test/printf-corners.ok _printf-corners differ: char 21, line 5
make[2]: [Makefile:4438: printf-corners] Error 1 (ignored)
.../test/printf-corners.ok _printf-corners differ: char 26, line 6
make[2]: [Makefile:4439: printf-corners] Error 1 (ignored)

 As verified on fencepost I am covered by the Red Hat copyright assignment 
with the FSF for contributions to GAWK, so this is fine to go in as it is.  
Sending from my personal e-mail address to avoid a technical issue with 
the Red Hat mail server interfering with patch sets.

 Please let me know if you have any questions, comments, or concerns, and 
I'll be happy to address them.

  Maciej