[Bug-tar] Fw: tar 1.27.1 compile error on AIX 7.1 TL2 SP3 and AIX 6.1 TL8 SP3 with IBM XL C/C++

[yaberger]

Hi Paul,

IBM support has been able to provide
me a workaround:
by replacing this:
  char minbuf[max (SYSINT_BUFSIZE,
INT_BUFSIZE_BOUND (intmax_t))];
  char maxbuf[max (SYSINT_BUFSIZE,
INT_BUFSIZE_BOUND (uintmax_t))];
by that:
  int minbufsize = max (SYSINT_BUFSIZE,
INT_BUFSIZE_BOUND (intmax_t));
  int maxbufsize = max (SYSINT_BUFSIZE,
INT_BUFSIZE_BOUND (uintmax_t));
  char minbuf[minbufsize];
  char maxbuf[maxbufsize];
and this:
      char buf[max (SYSINT_BUFSIZE,
INT_BUFSIZE_BOUND (intmax_t))];
by that:
      int bufsize = max
(SYSINT_BUFSIZE, INT_BUFSIZE_BOUND (intmax_t));
      char buf[bufsize];

I'm still waiting to see if they will
consider this as a compiler defect and if an APAR will be created to eventually.
I'll let you and the mailing list know
once I receive this answer.


Best regards,

Yannick Bergeron
Advisory IT Specialist
----- Forwarded by Yannick
Y Bergeron/Bromont/IBM on 01/06/2014 03:28 PM -----

From:      
 Yannick Y Bergeron/Bromont/IBM
To:      
 Paul Eggert <address@hidden>@IBMCA,

Cc:      
 address@hidden
Date:      
 12/26/2013 08:21 AM
Subject:    
   Re: [Bug-tar]
tar 1.27.1 compile error on AIX 7.1 TL2 SP3 and AIX 6.1 TL8 SP3 with IBM
XL C/C++

---

Hi Paul,

I've just submitted a ticket to IBM
XL C/C++ support team and will wait for their feedback.
Your support has been of a great help
and is really appreciated.
I'll let you know what is the result
of this ticket or when the problem will have been targetted.
Until then, have a good time.


Best regards,

Yannick Bergeron
Advisory IT Specialist



From:      
 Paul Eggert <address@hidden>
To:      
 Yannick Y Bergeron/Bromont/address@hidden,

Cc:      
 address@hidden
Date:      
 12/25/2013 06:15 PM
Subject:    
   Re: [Bug-tar]
tar 1.27.1 compile error on AIX 7.1 TL2 SP3 and AIX 6.1 TL8 SP3 with IBM
XL C/C++

---

Yannick Y Bergeron wrote:
> If it's a compiler issue, I can open a ticket to IBM XL C/C++ but
I would 
> need to understand why it should works (other than "it works
with gcc"), 
> what would be the normal behavior and why this should be considered
a 
> compiler issue.

Simply put, the program declares an identifier SYSINT_BUFSIZE,
but cc -qlanglvl=extc99 later incorrectly reports that this
identifier is not declared.

I'll attach a small sample.c file that attempts to take your
tar.output file and discard everything except what's needed
to illustrate the bug.  If your 'cc -qlanglvl=extc99 -g -c sample.c'
fails, the compiler is clearly buggy.  If cc accepts the program,
you may have to restore some of the stuff I've removed (indicated by
/* ... */) to make the bug reappear.  Either way, you can
use the program as part of your trouble ticket.  The simplest
thing for you would be to submit the full foo.c, which you've already
got, but it may help the IBM maintainers to have a smaller test
case that illustrates the bug.  I also suggest mentioning this email
thread

http://lists.gnu.org/archive/html/bug-tar/2013-12/msg00025.html

as part of your trouble ticket.

It would be helpful to come up with a way to work around the bug,
so that 'tar' can be built even with a buggy compiler.  But we may
need to know what the compiler bug is, before we can do that.
[attachment "sample.c" deleted by Yannick Y Bergeron/Bromont/IBM]