The fix for this is to use more ExceptionDispatchInfo! Basically
everwhere that we handle an exception is now an EDI. It's easy to pass
these around and they do the right thing as far as perserving the stack
in.
I de-factored this code a little bit to make all of this work, but it's
now pretty unambiguously correct upon inspection.
I started out wanting to get rid of the continue-with because we've
found that pretty error prone overall. However making this method
async-void started to ask more questions than it answered. What happens
if serialization of the exception fails?
\n\nCommit migrated from 0a27fd3ad6
* Remove obsolete targets, properties, and scripts
* Replace IsProductComponent with IsShipping
* Undo bad merge to version.props
* Update documentation, and put workarounds into a common file
* Replace usages of RepositoryRoot with RepoRoot
* Remove API baselines
* Remove unnecessary restore feeds and split workarounds into two files
* Enable PR checks on all branches, and disable autocancel\n\nCommit migrated from f41cfded3c
Nate McMaster
Larry Ewing
Ryan Nowak
Javier Calvarro Nelson
Pranav K
Chris Ross
Pavel Krymets
Ryan Brandenburg