Re: Gawk 'make check' on NSK/OSS

[Matthew Woehlke]

Aharon Robbins wrote:
> In article <address@hidden> you write:
> > On NSK/OSS, I get a couple of "minor" 'make check' failures followed by 
> > the fatal failure:
> > gawk: (FILENAME=./rsstart1.in FNR=1) fatal: error reading input file 
> > `./rsstart1.in': Invalid function argument
> > make: *** [rsstart1] Error 2
> >
> > Any pointers what the fatal error might be, or how to diagnose the 
> > "minor" failures? (Some of them don't look so minor - should I pack up 
> > the outputs and send them here?)
>
> Hi. You can run `make diffout' to capture the differences from what
> should be generated.

Ok, but what do I do with that? Some things look OK, like this:

============== _double1 =============
*** ./double1.ok        Tue Feb  6 11:52:47 2007
--- _double1    Fri Feb  9 12:19:35 2007
***************
*** 1,2 ****
  9.22337e+18
! 9223372036854775808
--- 1,2 ----
  9.22337e+18
! 9223372036854776000

...but what do I make of this?

============== _intformat =============
*** ./intformat.ok      Tue Jan 30 10:29:00 2007
--- _intformat  Fri Feb  9 12:17:51 2007
***************
*** 1,2 ****
--- 1,903 ----
+ (sprintf(0x%x,10^83) = 0x0)-7.45834e-72 = -7.45834e-72
+ (sprintf(%d,10^83) = 0)-7.45834e-72 = -7.45834e-72
+ (sprintf(%s,10^83) = 7.45834e-72)-7.45834e-72 = -9.80379e+75
+ (sprintf(%.0f,10^83) = 0)-7.45834e-72 = -7.45834e-72
... MANY lines of this sort of thing ...

Anyway, the failures(?) are double1, double2, fmtspcl, intformat, 
longsub, rsnulbig2, rsstart1 and rsstart2.

> As for the fatal error, you'll have to just step through it with a debugger
> and let me know what you see, unless I can get direct ssh access to the
> machine.

Um. There is such a thing as a debugger on NSK. It likes to pop up 
sometimes when something dies, but I have no idea how to use it or even 
how to start it. Any suggestions where I should start dropping printf()s 
in the source code?

> BTW, please use  bug-gawk AT gnu.org to make sure that a copy of your mail
> comes straight to my inbox. Otherwise, you're relying on my seeing stuff
> via USENET...

Sending to bug-gnu-utils is *not* the same as sending to bug-gawk? Hmm, 
I'll try to remember that...

--
Matthew
Congratulations! You've won a free trip to the future! All you have to 
do to claim your prize is wait five minutes...