Re: [htdig-dev] Copyright & LGPL Update + Native WIN32 Port checkin

[Jim C. <li...@yg...>]

On Thursday, June 26, 2003, at 11:24 AM, Neal Richter wrote:

> 	Please get your compilers humming and pound on this.  I made every
> effort to make sure the Win32 changes are transparent to non native 
> Win32
> builds.

Still seems to compile fine under OS X (as of Saturday's CVS). The 
'make check' tests pass, but seem to require some effort. I assume this 
is unrelated to your changes, but I will mention it here anyhow.

I ran into a number of mostly random failures that would appear on one 
run and go away on the next. I traced the problem to the startup of 
httpd in test_functions. There is no sleep after the call to start 
httpd, and this results in the server sometimes not being ready when 
dependent tests start making requests. After adding a 'sleep 2' (as 
test_functions already does whenever it kills httpd), I was unable to 
reproduce the failures.

There also appears to be another problem with test_functions involving 
the process used to kill httpd. The pattern seems to be to check for 
'log/httpd.pid', issue a 'kill' using the pid cat'd from the file if 
the file exists, and then copy '/dev/null' to 'logs/httpd.pid'. This 
leaves an empty file which later allows the test for 'log/httpd.pid' to 
succeed when there is in fact no pid on which 'kill' can operate. The 
result is that 'kill' is called without an argument. This does not 
typically interfere with the actual tests, but does generate noise 
might concern people who run the tests. This problem also appears to be 
someone random. The test(s) on which the 'kill' is called without 
arguments varies; and in some cases the problem doesn't occur at all.

Jim