Run CI on mac os only with go1.13 and on windows only on go1.14.
Speed up tests. Introduce --allow-parallel-runners.
Block on parallel run lock 5s instead of 60s.
Don't invalidate analysis cache for minor config changes.
The primary improvement is in early clearing of
analyzed package's TypeInfo, facts, etc for
whole program analyzers (`unused`). Clear it when it
becomes unused and GC collects them early. Initially this
clearing was performed for all analyzers except `unused`.
Update staticcheck from v0.0.1-2019.2.3 to v0.0.1-2020.1.4
Also in this commit:
* speed up loading packages from export data (2.5s -> 2.1s for std)
by not using mutex for export data since it was allowed in
x/tools#07722704da13
* make an order of execution of linters stable
* update renameio and robustio
* use robustio in caching
Relates: #987, #994, #995, #1011
* update staticcheck
Don't fork staticcheck: use the upstream version.
Remove unneeded SSA loading.
* Cache go/analysis facts
Don't load unneeded packages for go/analysis.
Repeated run of go/analysis linters now 10x faster
(2s vs 20s on this repo) than before.