Testing for the 4.4p1 release

Darren Tucker dtucker at zip.com.au
Sat Sep 2 13:03:24 EST 2006

Hi, and thanks for testing.

Iain Morgan wrote:
> Sometime ago, Iain Morgan wrote:
>> The 20060830 snapshot configures, but fails to build on IRIX 6.5.29
>> usign the MIPSpro 7.4 compilers.
> Adding <errno.h> to the headers included in port-irix.c allows the build
> to complete.

This has been added, thanks.

> Using c99, there are only 92 warnings, whereas using cc there are 251.
> A reasonable number of the 92 warnings are due to macro redefinition:
>  14   Macro "_PATH_MAILDIR" (declared at line 359 of "defines.h") has an
>   14   Macro "_PATH_BSHELL" (declared at line 322 of "defines.h") has an
> incompatible

These are defined in /usr/include/paths.h, right?  I think we should put 
that back into includes.h or maybe defines.h, because the ones we 
default to in defines.h may not be right.

>    5   Macro "offsetof" (declared at line 491 of "defines.h") has an
> incompatible
>    1   Macro "offsetof" (declared at line 491 of "../defines.h") has an
> incompatible

Which system header defines that?

> Also, moving -lgen from LIBS to SSHDLIBS would eliminate 10 additional
> warnings.

If it's not a critical problem I'd prefer to wait until after the 
release to do any library reshuffles, but there is definitely some room 
for improvement.

> I haven't had time to deal with the regression tests yet. The protocol 1
> tests always seem to be a problem with IRIX. (I reported this once quite
> a while ago, but did not have the time to investigate.) In the past,
> I've hacked the scripts to only test protocol 2, in which case
> everything (that we care about) is fine.

The only time I've ever seen protocol 1 specifically fail regress tests 
is on Solaris 10 with the bundled OpenSSL, and that's because any crypto 
with a key length >128bit had been crippled including the blowfish 
cipher that protocol 1 uses.  (It also included the AES functions, but 
OpenSSH has its own implementation of those and now uses them if it 
finds the native ones don't work.)

When I'm trying to debug regress tests (which, I'll admit, isn't easy) I 
do something like the attached patch, which stops immediately after the 
failure without cleaning up.  I then rerun the failing command with 
extra debugging (usually ssh -vvv [lots of args]) which can help shed 
some light on the failure.

Darren Tucker (dtucker at zip.com.au)
GPG key 8FF4FA69 / D9A3 86E9 7EEE AF4B B2D4  37C9 C982 80C7 8FF4 FA69
     Good judgement comes with experience. Unfortunately, the experience
usually comes from bad judgement.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: openssh-regress-debug.patch
Url: http://lists.mindrot.org/pipermail/openssh-unix-dev/attachments/20060902/d9a8ba1b/attachment.ksh 

More information about the openssh-unix-dev mailing list