Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Conversation

@kslr
Copy link
Contributor

@kslr kslr commented Oct 28, 2019

Reverts #1893

@kslr kslr merged commit e55f1c5 into master Oct 28, 2019
@lixin9311
Copy link
Contributor

I can't see any logic to revert it.
Windows implementation behavior will be inconsistent with Other implementations now.
The issue mentioned in #1982 was caused by wrong configuration, not the patch.
He must sign his certificate correctly to establish TLS connections, not by ignoring certificate issues

@vcptr
Copy link
Contributor

vcptr commented Dec 19, 2019

What mentioned at #1982 is that PR #1893 causes the windows clients not trusting LetsEncrypt certs. Thus we revert it to the original way. The PR has side effects.

By calling loadSelfCertPool, it overrides the system certs on windows. That's why raymond introduced the DisableSystemRoot config.

Maybe there are better ways handling the situation, but simply ignore the DisableSystemRoot mechanism surely not.

@vcptr
Copy link
Contributor

vcptr commented Dec 19, 2019

By investigating deeper on the cert loading code (sorry, my first time digging into them), I found that on windows, x509.SystemCertPool() is not available.

I agreed what you said, weather DisableSystemRoot on or off, core should load the custom certs, but on windows you can not combine system certs and custom certs easily.

The inconsistency of behavior is an OS level problem, I believe. I also see that their's work-arounds getting the system certs on win, but it seems requiring administrator permissions, and it's complicated, I don't like it.

@nicholascw nicholascw deleted the revert-1893-patch-1 branch June 1, 2020 12:41
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants