tree ecb92171bdfb4798f3613fce5d92e0c33b0cc7ee
parent 4df57136e9832327c11e48b5bfe00b0326bd5b63
author ocornut <omarcornut@gmail.com> 1613144062 +0100
committer ocornut <omarcornut@gmail.com> 1613144062 +0100

Made a change to CalcWindowSizeAfterConstraint() which MSVC optimizer seems to trip on. (read on)

We seldomly modify a local ImVec2 passed as parameter. This should be perfectly legal but libigl stumbled on a what may be a MSVC optimizer bug?
Only seven function in the codebase modify as 'ImVec2 parameter' but this is the only static one (and fairly small that is) which may tempt optimizer to perform optimizations over caller-callee boundaries?
While we can't explain it fully yet and I assume it may haunt us back someday. Might also be chain reaction from some unobvious UB but can't see it right now...
Link: https://github.com/libigl/libigl/issues/1669