Closed isedev closed 3 years ago
Added the new fix. Let me know what you think.
cc @emersion @mschneider82 @TimWolla @jefferai (I wish I remembered everyone's name/handle). Someone at Snyk is also willing to help getting a CVE for this which will be a first for me. Thanks everybody out there paying attention and willing to help.
Thanks for the ping! I'm afraid I won't be much of a help here, though. I'm not even a go programmer. My contribution regarding the unique ID TLV was motivated by the fact that I was the person that added this TLV to the PROXYv2 spec (I'm somewhat involved with upstream HAProxy) and I just searched for proxy protocol libraries to make sure they are up to date.
@pires Thanks for pinging me! This generally looks good, left a couple of comments.
The code mostly LGTM. Other than two comments I kindly request you tidy up the commit log.
Any suggestions how I tidy up the commit log? I agree needs to be done, just no idea how to without new pull request. If that's what it takes, more than happy to.
Any suggestions how I tidy up the commit log?
Use an interactive rebase git rebase -i
and then force push to isedev:feature/issue-69-tests
.
Any suggestions how I tidy up the commit log?
Use an interactive rebase
git rebase -i
and then force push toisedev:feature/issue-69-tests
.
Thanks. Learn something new everyday :)
@pires can you tag a new release?
Working on it. Want to get Go 1.16 in and some fixes to tests I spotted while doing so.
i do a pull req for traefik, where i used this lib to provide proxy protocol support, but traefik still depends on go 1.15, so requiring 1.16 here would break it. I will stick on last commit for now
There's no such thing as enforcing Go 1.16 requirement. CI compiles and tests for Go 1.14 and Go 1.15, and now want to add 1.16, that's all.
Ok fine. The traefik maintainers requested a new version tag
OK the tests fixing will have to wait as I need more time to find a reasonable strategy while also refactoring to reduce lots of duplicated code.
2As suggested in PR#70, a new pull request which captures just the tests proving issue #69, with some refinements:
The second issue above can be used for DoS by consuming server process sockets.
Will then submit fixes - further testing shows that pull request #70 does not solve the problem properly.
Fixes #69