Re: pkg/48566 (getpwnam() fix and ECDH for sendmail from upstream)

[jnemeth%CornerstoneService.ca@localhost (John Nemeth)]

The following reply was made to PR pkg/48566; it has been noted by GNATS.

From: jnemeth%CornerstoneService.ca@localhost (John Nemeth)
To: Emmanuel Dreyfus <manu%netbsd.org@localhost>, 
gnats-bugs%netbsd.org@localhost
Cc: gnats-admin%netbsd.org@localhost, pkgsrc-bugs%netbsd.org@localhost
Subject: Re: pkg/48566 (getpwnam() fix and ECDH for sendmail from upstream)
Date: Wed, 18 Jun 2014 22:06:55 -0700

 On Jun 17,  3:39pm, Emmanuel Dreyfus wrote:
 } On Tue, Jun 17, 2014 at 02:15:00PM +0000, Emmanuel Dreyfus wrote:
 } >  And with -current pkgsrc/mail/sendmail/paches, I can restore 
functionnality 
 } >  with this simple change (and make makepatchsum of course), hence 
re-entrant
 } >  resolver is the culprit.
 } 
 } I found two bugs in the pkgsrc patches for NAMED_RESN
 } - res_ninit is called regardless if the structure was already initialized
 
      The case where res_ninit() was called without checking for
 previous initialisation was relatively early in the startup.  A
 visual code inspection indicated that it was highly unlikely that
 res_ninit() would be called after initialisation.  Did you find a
 code path where it could happen?  I didn't bother testing for
 initialisation at that point because it made things rather awkward.
 However, just to make you happy I figured out a way to do it.  Note
 that in the unlikely event that res_ninit() was called when it had
 already been called, all that would have happened is a small amount
 of leaked memory.  Not great, but hardly the end of the world.
 BTW, even in the case that res_init()/res_ninit() didn't get called,
 the code went ahead and initialised all the options anyways.  Which
 is kind of silly, but not a change created by this patch.
 
 } - there is a res_search left in sendmail/sm_resolve.c that does not have
 
      Thank you for pointing that out.  It was the only place that
 got missed.  I found it and squashed it.  I have also verified with
 the help of nm that all deprecated functions, as listed in resolver(3),
 have been found and eradicated.
 
 } a res_nsearch counterpart. If NAMED_RESN is defined I think it will operate
 } without proper resolver initiliazation.
 
      BTW, where were you when I asked for testing and feedback
 prior to 8.14.9 being committed?
 
 }-- End of excerpt from Emmanuel Dreyfus