Re: Retiring WITHOUT_CXX
- In reply to: Rodney W. Grimes: "Re: Retiring WITHOUT_CXX"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Fri, 26 Nov 2021 18:48:44 UTC
On Fri, Nov 26, 2021 at 11:26 AM Rodney W. Grimes < firstname.lastname@example.org> wrote: > > -------- > > Baptiste Daroussin writes: > On Fri, Nov 26, 2021 at 09:09:54AM -0800, > Rodney W. Grimes wrote: > > > > > > You can disagree with my assertion, but I shall continue to assert > > > > that it *seems* as if rather than adding B O S to the CI > > > > I dont know how long time B-O-S runs these days, it used to take > > a couple of weeks, and I have a hard time squaring that with any > > contemporary use of the "CI" paradigme :-) > > Dexter can provide an exact number, but IIRC on his purpose built > test box it is on the order of 2 or 3 days. > I can do a clean buildworld in about ~20-30 minutes on my fastest machine. There's ~300 build options which puts us at ~74 hours, or ~3 days. I've not done a build survey, though, to know for sure. Note that this doesn't do combinations, which is a combinatoric explosion in size (even doing all pairs would be 150x slower). Which is at least ~20x too slow to put into any pre-commit workflow, but would be suitable for periodic testing, perhaps. But then we need to ask if that ~75 hours of CPU time on a 128-thread CPU generates useful information in proportion to the size of the test... But it boils down to all things having a cost. And some costs are too great. Warner