Fix repeated conflict detection and introduce optional resolver timeout

[viktor-svub]

• repeated conflict detection used HashSet as set item -- as it was not comparable, it was always adding new records -- it was replaced by F# Set which implements comparable as expected
• added support for PAKET_RESOLVER_TIMEOUT environment variable, allowing to terminate the resolver in finite time -- for cases when we hit "loop of doom" on automated build agents
• switched to UTC time handling to avoid DST glitches (plus, UtcNow is faster)

[forki]

oups ;-)

[matthid]

Good, maybe that explains our changing Lockfile as well.

[forki]

ok there are two integration tests failing. I think it's because the output changed

[viktor-svub]

I may be mistaken but it looks to me like the failure is unrelated -- i.e. the #1177 should resolve with pessimistic strategy correctly failed with:

-> Could not find versions for package Castle.Windsor on http://www.nuget.org/api/v2.
-> Source 'http://www.nuget.org/api/v2' yielded no results
	-  Request 'http://www.nuget.org/api/v2/FindPackagesById()?semVerLevel=2.0.0&id='Castle.Windsor'' finished with error
	-> Could not retrieve data from 'http://www.nuget.org/api/v2/FindPackagesById()?semVerLevel=2.0.0&id='Castle.Windsor''
	-> HttpRequestException: An error occurred while sending the request
	-> WebException: Error: TrustFailure (One or more errors occurred.)
	-> AuthenticationException: A call to SSPI failed, see inner exception.
	-> MonoBtlsException: Ssl error:1000007d:SSL routines:OPENSSL_internal:CERTIFICATE_VERIFY_FAILED
	     at /build/mono-5.4.1.7/external/boringssl/ssl/handshake_client.c:1132

And #55 should resolve with pessimistic strategy correctly is similar, just with different package.
Isn't this something changed recently on master?

[forki]

master is green AFAIK

[viktor-svub]

Also locally (Windows only) the build RunIntegrationTestsNetCore SkipDocs=1 with PAKET_RESOLVER_WORKERS=24 and PAKET_TESTSUITE_FLAKYTESTS=false passed all tests.

Right, I have no Mono install, and master is green, but I cannot see how these changes affected certificate validation against nuget.org -- maybe this PR is not re/based correctly?

[forki]

ok let me try to rebase it in new PR

[forki]

let's see #3085

[forki]

the other one is green

[forki]

Thx. It's released