Kohei Yoshida <[EMAIL PROTECTED]> writes:

> 1) The use of "product" and "non-product" terms seems unclear to me.
> What do they mean exactly?
> 
Hi Kohei,

a "product" version is one that has .pro suffix at the output trees
(that's the default case). A "non-product" has no such suffix, and
gets enabled via --enable-dbgutil at the configure line. We should
rather advertise this a bit more, because people then get assertions
if they break stuff horribly. ;-)

> 2) IMO, requiring that the developer of the cws make the binary install
> set available to the QA personnel has the following downside.
> 
> On Linux platform, there is an issue of ABI compatibility due to gcc
> versioning as well as system library dependencies.  When I build on my
> machine, I do build using gcc 4.1.0 and make use of external system
> libraries.  This means that, even if I am able to provide an
> installation set for the QA personnel by uploading it to an FTP/Web
> server, my install set may not run on QA person's (Linux) machine.
> 
> To me, it is just as well workable (or better?) to check the integrity
> of a cws at source code level, and have both the developer and the QA
> build the same cws on both ends.  I'm personally not seeing any
> advantage of requiring the developer to build and provide the binary
> install sets for QA, especially on Linux platform.
> 
Yep, sounds reasonable. Would need some kind of automatism on our
side, to enable more people to trigger builds. But that appears to be
just another case for a (specialized) buildbot...

Cheers,

-- Thorsten

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to