Did anybody else noticed an extreme build speed regression (not test execution) in the latest NCrunch version? I just did some rough testing and I've got following result:
Time needed to build and execute 3 test cases after modifying single line of code in test file (add new line)
NCrunch 2.16.0.13
~3 sec
NCrunch 2.17.0.16
~15 sec. ~13 sec in build phase
My Environment:
Visual Studio 2015
Solution Contains 37 projects, 15 of them are Test Projects (MS Test)
NCrunch Engine Mode: Run impacted tests
You are the first person to report this problem, but that doesn't mean it isn't common.
Do you have any other solutions you can test the performance on? It would be very useful if we can identify whether there is something solution-specific that is surfacing the problem, or whether this is more general.
A 3sec to 15sec is a serious regression and is not intentional.
You are the first person to report this problem, but that doesn't mean it isn't common.
Do you have any other solutions you can test the performance on? It would be very useful if we can identify whether there is something solution-specific that is surfacing the problem, or whether this is more general.
A 3sec to 15sec is a serious regression and is not intentional.
No, unfortunately I do not have a "test solution" that replicates the problem. I'll try cook something up, can't promise anything though.
The 2.17 release contained a change that disabled shared compiler usage during build. Early (pre-release) versions of VS2015 had some stability problems with this feature, and it didn't seem to be needed for NCrunch, so the sensible thing seemed to be to turn it off. Some tests recently done on my side suggest that this seems to have had an impact on build speed. I've re-enabled shared compiler usage for the above build, so I'd be interested in hearing how it goes for you.
The 2.17 release contained a change that disabled shared compiler usage during build. Early (pre-release) versions of VS2015 had some stability problems with this feature, and it didn't seem to be needed for NCrunch, so the sensible thing seemed to be to turn it off. Some tests recently done on my side suggest that this seems to have had an impact on build speed. I've re-enabled shared compiler usage for the above build, so I'd be interested in hearing how it goes for you.