You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
To sumarize the problem, there are SSL tests failing 99% of the time on Linux VMs (amd64, Github Action). They never fail on my Mac (arm64) and fail sometimes when ran in docker (arm64, linux/amd64).
You can trigger the error by running bundle exec cucumber features/ftp_server/implicit_tls.feature, the result is always:
550 Connection reset by peer
(Net::FTPPermError)
I found the issue goes away when I remove the changes from f55e9d5.
As far as I can tell, overriding rbuf_fill when shutdown has been called, prevents the current exchange from finishing properly.
I'm not sure if this change is still relevant today, there was no tests so I'm not even sure what it was supposed to fix exactly.
I found socketry/async-io#61, referencing ruby/openssl#609, and based on the changes applied, I tried to replace the rbuf_fill override with a simpler change in shutdown:
defshutdown(*args)@io.__send__(:stop)end
This fixes the issue and seems to allow the tests to be fast (as opposed to no doing anything in shutdown for example).
Do you find this change reasonable? If you do, then I can submit a PR for it.
The text was updated successfully, but these errors were encountered:
Hi, I'm currently trying to update ftpd to work with recent ruby versions, and I'm having a hard time having the tests pass.
You can see my changes in wconrad/ftpd#59.
To sumarize the problem, there are SSL tests failing 99% of the time on Linux VMs (amd64, Github Action). They never fail on my Mac (arm64) and fail sometimes when ran in docker (arm64, linux/amd64).
You can trigger the error by running
bundle exec cucumber features/ftp_server/implicit_tls.feature
, the result is always:I found the issue goes away when I remove the changes from f55e9d5.
As far as I can tell, overriding
rbuf_fill
when shutdown has been called, prevents the current exchange from finishing properly.I'm not sure if this change is still relevant today, there was no tests so I'm not even sure what it was supposed to fix exactly.
I found socketry/async-io#61, referencing ruby/openssl#609, and based on the changes applied, I tried to replace the
rbuf_fill
override with a simpler change inshutdown
:This fixes the issue and seems to allow the tests to be fast (as opposed to no doing anything in
shutdown
for example).Do you find this change reasonable? If you do, then I can submit a PR for it.
The text was updated successfully, but these errors were encountered: