Michel Hidalgo
Michel Hidalgo
This one's still causing regressions (I think): https://ci.ros2.org/view/nightly/job/nightly_win_rel/lastCompletedBuild/testReport/projectroot/test/test_timer__rmw_connextdds/
@bpwilcox are you planning to circle back on this?
@bpwilcox do you still plan on submitting changes? Or shall I close this patch?
@bpwilcox friendly ping
IMO our main issue here is that there's no clear path forward. Even if we decide to rely on all libraries sharing the same error handling state, `rcutils` error handling...
FYI @Blast545 @brawner.
> Thought a little on this, what if a "build" flag is set then the overwritten errors are print to stderr and if not set the error is overwritten without...
That's one way forward, but I will point out that: - Propagating error states with the current API is clunky. Namely, chaining error states properly is hard, and we don't...
> Chunkiness can be fixed with a better API. Sure. I was just stating that if we go that way we probably have to roll out new API. > Is...
> I don't get this point, my suggestion was to not build backtraces but to handle existing errors properly, only combining their message text, as needed, into a single error...