py27-omniORBpy pre-built binary trouble

[Ryan Schmidt]

On Sep 8, 2014, at 2:35 PM, Thomas G Lockhart wrote:
> 
> On Sep 8, 2014, at 1:05 AM, Ryan Schmidt wrote:
> 
>> 
>> On Sep 8, 2014, at 12:37 AM, Thomas G Lockhart wrote:
>>> 
>>> Great. btw, it seems that the patch files are no longer necessary, though I have not yet done enough interoperability testing to be absolutely certain. I’ll open a ticket at some point to get them to go away.
>> 
>> I agree the patch for CORBA_sysdep.h is no longer necessary; it was integrated upstream, but it just so happened that the patch was still able to apply again, with no harm done. I've removed it in r125162.
> 
> Great.
> 
>> The patch for config-darwin.h needs further scrutiny. In fact the definitions in that file look very suspicious and could be the reason why the universal build would fail.
> 
> The build system for omniORB has a lot of history and more recently has autoconf blended in on top of it. I find it pretty impenetrable and would be surprised if we can get a universal build working.

Actually when I tried again, the normal universal build succeeded.


> One issue which might make this harder is that the CORBA spec requires specific widths for the integer data types. The implementation probably uses #define constructs to force the right widths depending on the architecture, so having the same code base and just flipping compiler flags may not be enough.
> 
> I guess that the universal build was claimed to work in the past. But I have no idea if there was any test coverage to verify that the CORBA on-the-wire data widths were in fact correct. I’m skeptical about the whole thing if you haven’t already deduced this ;)

I am skeptical about it too and would want to know whether it was actually working. muniversal might be safer. I'll see if I can wire that up, after fixing the config file.