Re: 0.05 Update [was Re: 0.05]

[Brian Jones]

Julian Dolby <address@hidden> writes:

> RedHat 7.3 and 8.0 systems the man page provided for `man 7 socket' claims
> the following:
> 
>        SO_RCVTIMEO and SO_SNDTIMEO
>               Specify the sending  or  receiving  timeouts  until
>               reporting  an  error.  They are fixed to a protocol
>               specific setting in Linux and  cannot  be  read  or
>               written.  Their functionality can be emulated using
>               alarm(2) or setitimer(2).
> 
>  Does `man 7 socket' say something different on your Debian system? That
> was why I disabled those options.  As I recall, the code compiled fine
> under Linux but threw errors when it was actually used with the timeout
> option.  I will try Eclipse on Jikes RVM without this patch, and let you
> know what happens.

The test in Mauve seems to bare this out, an exception is thrown in
our native code.  This led me to discovering I had to wait several
minutes between test runs due to SO_REUSEADDR not being set on the
ServerSocket.  Looks like gcj does this by default during 'bind', but
I can't see in the docs where this is how it should be.  I added it to
the test helper class and found it didn't work (not supported in
native code).  So I added it.  Now my question, it doesn't appear that
our native code for these options matches the constant values
expressed in java/net/SocketOptions, so should I create some dummy
defines in javanet.c that use the correct values with things like

#define GCP_SO_REUSEADDR 4
...
case GCP_SO_REUSEADDR:
  setsockopt (..., SO_REUSEADDR, ...);
...

Brian
-- 
Brian Jones <address@hidden>