[CMake] 2.8.1 vs 2.8.0

Phil Smith phil at voltage.com
Thu Apr 1 16:10:15 EDT 2010


Those who have had my dumb questions inflicted on them in the past will recall that I'm using CMake with a cross-compiler (Dignus) for IBM System z (mainframes).

We recently added a new developer, and in setting her up, we had problems with:
-- Check size of unsigned long long - failed
(and all the other "-- Check size of ...").

These fail quietly, with no compile errors. Past experience suggests that this is due to the regex not finding the magic string in the compiler output.

After some tinkering, my user backed off to CMake 2.6, and now it works.

I'm running 2.8.0 without problems; it looks like the process for doing the "check size" stuff changed in 2.8.1.

One thing that was a huge impediment to figuring this out before was that the object created by the "check size" process seems to get deleted even if the process fails. That seems wrong: shouldn't it leave the droppings for diagnosis if it fails?

Anyway, any hints about what this change means in terms of how it processes stuff, and where to look to figure out the issue? Fortunately I have a sacrificial machine that I can test on.
-- 
...phsiii

Phil Smith III
phil at voltage.com
Voltage Security, Inc.
www.voltage.com



More information about the CMake mailing list