I don't get it.  Neither the NS-3 library nor the NS-3 Python module use any
kind of '_environ' symbol.

Moreover I am no longer able to ssh into darwin-ppc like I used to, so I am not
able to test this.

Created an attachment (id=311) [details]
darwin-ppc build fix

This patch fixes build errors on the darwin-ppc regression host.  It builds
fine there and runs Python unit tests.  Still no idea about this _environ issue
reported here.

(From update of attachment 311 [details])
I am waiting for approval from whoever owns the icmpv4 code to commit this.

(In reply to comment #2)
> Created an attachment (id=311) [details] [details]
> darwin-ppc build fix
> 
> This patch fixes build errors on the darwin-ppc regression host.  It builds
> fine there and runs Python unit tests.  Still no idea about this _environ issue
> reported here.
> 

I really don't like this patch: the code should be fine without it. Would you
mind show me the compiler error message together with compiler version ?

(In reply to comment #4)
> (In reply to comment #2)
> > Created an attachment (id=311) [details] [details] [details]
> > darwin-ppc build fix
> > 
> > This patch fixes build errors on the darwin-ppc regression host.  It builds
> > fine there and runs Python unit tests.  Still no idea about this _environ issue
> > reported here.
> > 
> 
> I really don't like this patch: the code should be fine without it. Would you
> mind show me the compiler error message together with compiler version ?

The bug happens because the python headers include /usr/include/sys/termios.h
which defines the macro ECHO to 0x000008 

This is really butt-ugly, especially since if you grep for ICMP_ECHO in
/usr/include/*, you will find netinet/ip_icmp.h which defines it so, adding the
ICMP prefix is not bullet-proof.

I don't really know what to do: the C preprocessor is a killer.

(In reply to comment #5)
> The bug happens because the python headers include /usr/include/sys/termios.h
> which defines the macro ECHO to 0x000008 
> 
> This is really butt-ugly, especially since if you grep for ICMP_ECHO in
> /usr/include/*, you will find netinet/ip_icmp.h which defines it so, adding the
> ICMP prefix is not bullet-proof.

Similar approach was followed by e.g. SocketError.

> 
> I don't really know what to do: the C preprocessor is a killer.

If it were up to me I would still commit this patch as is.  It may not be
bullet proof, but it fixes our immediate problem.  I don't like fixing
potential/unlikely problems.

Another approach would be to #ifdef ECHO #undef ECHO #endif, after including
<Python.h> but before including ns-3 headers.

(In reply to comment #6)

> Another approach would be to #ifdef ECHO #undef ECHO #endif, after including
> <Python.h> but before including ns-3 headers.

If that works, I would very much prefer this.

(In reply to comment #7)
> (In reply to comment #6)
> 
> > Another approach would be to #ifdef ECHO #undef ECHO #endif, after including
> > <Python.h> but before including ns-3 headers.
> 
> If that works, I would very much prefer this.
> 

In that case, let's not argue about it.  Committed.

Only problem remaining is the _environ issue, which I cannot reproduce on
darwin-ppc unfortunately.

(In reply to comment #8)
> (In reply to comment #7)
> > (In reply to comment #6)
> > 
> > > Another approach would be to #ifdef ECHO #undef ECHO #endif, after including
> > > <Python.h> but before including ns-3 headers.
> > 
> > If that works, I would very much prefer this.
> > 
> 
> In that case, let's not argue about it.  Committed.
> 
> Only problem remaining is the _environ issue, which I cannot reproduce on
> darwin-ppc unfortunately.
> 

I cannot reproduce the _environ issue anymore either, so please mark as fixed
if it passes the overnight tests.