@sxa

Of course, though ... the only way to test ... is to build it without ccache and compare ... at which point you've already redone the build. :)

Though it would be interesting to do such testing systematically on an ongoing basis and nudge ccache towards the expectation of bit-for-bit reproducibility, at least under certain conditions.

I poked at this sort of thing ages ago, and recall the main issues were, surprise, timestamps, and possibly build path normalization...