Re: lib/41085: sigaction returns EINVAL instead of -1

[Matteo Beccati <matteo%beccati.com@localhost>]

The following reply was made to PR lib/41085; it has been noted by GNATS.

From: Matteo Beccati <matteo%beccati.com@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: lib/41085: sigaction returns EINVAL instead of -1
Date: Sat, 28 Mar 2009 01:08:21 +0100

 > From: David Holland <dholland-bugs%netbsd.org@localhost>
 >  On Fri, Mar 27, 2009 at 01:15:01PM +0000, matteo%beccati.com@localhost 
 > wrote:
 >   > 43              retval = sigaction(signo, &act, &oact);
 >   > (gdb) step
 >   > 45              if (retval < 0)
 >   > (gdb) print retval
 >   > $1 = 22
 >  
 >  That's bizarre. If this is repeatable, can you try running it under
 >  ktrace? That would reveal at least which sigaction call it's actually
 >  making, and if the return/error value shows up strangely under ktrace
 >  it will be a big hint.
 
 Even more bizzare is the fact that I can't see any call to the 
 __sigaction_sigtramp when sigaction() is returning 22:
 
 Proper call to sigaction(SIGALRM), with PHP outputting the result bool(true)
 
   15993      1 php      RET   write 12/0xc
   15993      1 php      CALL 
 __sigaction_sigtramp(SIGALRM,0xbfbfcfbc,0xbfbfcfa4,0xbaf58e6c,2)
   15993      1 php      RET   __sigaction_sigtramp 0
   15993      1 php      CALL  write(1,0x83b30f4,0xb)
   15993      1 php      GIO   fd 1 wrote 11 bytes
         "bool(true)
         "
 
 Call with a wrong callback, PHP outputs an error, plus the false result:
 
   15993      1 php      RET   write 11/0xb
   15993      1 php      CALL  write(1,0x83b3254,0x8c)
   15993      1 php      GIO   fd 1 wrote 140 bytes
         "
          Warning: pcntl_signal(): Invalid value for handle argument 
 specified i\
          n /root/compile/php-HEAD/ext/pcntl/tests/pcntl_signal.php on 
 line 10
         "
   15993      1 php      RET   write 140/0x8c
   15993      1 php      CALL  write(1,0x83b30f4,0xc)
   15993      1 php      GIO   fd 1 wrote 12 bytes
         "bool(false)
         "
 
 This is the interesting bit. PHP outputs bool(true) because sigaction 
 returned 22. No trace of the sigaction(-1) call.
 
   15993      1 php      RET   write 12/0xc
   15993      1 php      CALL  write(1,0x83b30f4,0xb)
   15993      1 php      GIO   fd 1 wrote 11 bytes
         "bool(true)
         "
 
 It seems that sigaction() validates signo itself and returns EINVAL 
 without making the syscall, or something along the lines.