Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Rollup of 5 pull requests #73615

Closed
wants to merge 15 commits into from

Conversation

Dylan-DPC-zz
Copy link

Successful merges:

Failed merges:

r? @ghost

Carsten Andrich and others added 15 commits May 1, 2020 17:52
add Windows system error codes that should map to io::ErrorKind::TimedOut

closes rust-lang#71646

**Disclaimer:** The author of this pull request has a negligible amount of experience (i.e., kinda zero) with the Windows API. This PR should _definitely_ be reviewed by someone familiar with the API and its error handling.

While porting POSIX software using serial ports to Windows, I found that for many Windows system error codes, an `io::Error` created via `io::Error::from_raw_os_error()` or `io::Error::last_os_error()` is not `io::ErrorKind::TimedOut`. For example, when a (non-overlapped) write to a COM port via [`WriteFile()`](https://docs.microsoft.com/en-us/windows/win32/api/fileapi/nf-fileapi-readfile) times out, [`GetLastError()`](https://docs.microsoft.com/en-us/windows/win32/api/errhandlingapi/nf-errhandlingapi-getlasterror) returns `ERROR_SEM_TIMEOUT` ([121](https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--0-499-)). However, an `io::Error` created from this error code will have `io::ErrorKind::Other`.

Currently, only the error codes `ERROR_OPERATION_ABORTED` and `WSAETIMEDOUT` will instantiate `io::Error`s with kind `io::ErrorKind::TimedOut`.
This makes `io::Error::last_os_error()` unsuitable for error handling of syscalls that could time out, because timeouts can not be caught by matching the error's kind against `io::ErrorKind::TimedOut`.

Downloading the [list of Windows system error codes](https://gist.github.com/carstenandrich/c331d557520b8a0e7f44689ca257f805) and grepping anything that sounds like a timeout (`egrep -i "timed?.?(out|limit)"`), I've identified the following error codes that should also have `io::ErrorKind::TimedOut`, because they could be I/O-related:

Name | Code | Description
--- | --- | ---
`ERROR_SEM_TIMEOUT` | [121](https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--0-499-) | The semaphore timeout period has expired.
`WAIT_TIMEOUT` | [258](https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--0-499-) | The wait operation timed out.
`ERROR_DRIVER_CANCEL_TIMEOUT` | [594](https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--500-999-) | The driver %hs failed to complete a cancelled I/O request in the allotted time.
`ERROR_COUNTER_TIMEOUT` | [1121](https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--1000-1299-) | A serial I/O operation completed because the timeout period expired. The IOCTL_SERIAL_XOFF_COUNTER did not reach zero.)
`ERROR_TIMEOUT` | [1460](https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--1300-1699-) | This operation returned because the timeout period expired.
`ERROR_CTX_MODEM_RESPONSE_TIMEOUT` | [7012](https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--6000-8199-) | The modem did not respond to the command sent to it. Verify that the modem is properly cabled and powered on.
`ERROR_CTX_CLIENT_QUERY_TIMEOUT` | [7040](https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--6000-8199-) | The client failed to respond to the server connect message.
`ERROR_DS_TIMELIMIT_EXCEEDED` | [8226](https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--8200-8999-) | The time limit for this request was exceeded.
`DNS_ERROR_RECORD_TIMED_OUT` | [9705](https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--9000-11999-) | DNS record timed out.
`ERROR_IPSEC_IKE_TIMED_OUT` | [13805](https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--12000-15999-) | Negotiation timed out.

The following errors are also timeouts, but they don't seem to be directly related to I/O or network operations:

Name | Code | Description
--- | --- | ---
`ERROR_SERVICE_REQUEST_TIMEOUT` | [1053](https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--1000-1299-) | The service did not respond to the start or control request in a timely fashion.
`ERROR_RESOURCE_CALL_TIMED_OUT` | [5910](https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--4000-5999-) | The call to the cluster resource DLL timed out.
`FRS_ERR_SYSVOL_POPULATE_TIMEOUT` | [8014](https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--6000-8199-) | The file replication service cannot populate the system volume because of an internal timeout. The event log may have more information.
`ERROR_RUNLEVEL_SWITCH_TIMEOUT` | [15402](https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--12000-15999-) | The requested run level switch cannot be completed successfully since one or more services will not stop or restart within the specified timeout.
`ERROR_RUNLEVEL_SWITCH_AGENT_TIMEOUT` | [15403](https://docs.microsoft.com/en-us/windows/win32/debug/system-error-codes--12000-15999-) | A run level switch agent did not respond within the specified timeout.

Please note that `ERROR_SEM_TIMEOUT` is the only timeout error I have [seen in action](https://gist.github.com/carstenandrich/10b3962fa1abc9e50816b6460010900b). The remainder of the error codes listed above is based purely on reading documentation.

This pull request adds all of the errors listed in both tables, but I'm not sure whether adding all of them makes sense. Someone with actual Windows API experience should decide that.

I expect these changes to be fairly backwards compatible, because only the error's [`.kind()`](https://doc.rust-lang.org/std/io/struct.Error.html#method.kind) will change, but matching the error's code via [`.raw_os_error()`](https://doc.rust-lang.org/std/io/struct.Error.html#method.raw_os_error) will not be affected.
However, code expecting these errors to be `io::ErrorKind::Other` would break. Even though I personally do not think such an implementation would make sense, after all the docs say that `io::ErrorKind` is _intended to grow over time_, a residual risk remains, of course. I took the liberty to ammend the docstring of `io::ErrorKind::Other` with a remark that discourages matching against it.

As per the contributing guidelines I'm adding @steveklabnik due to the changed documentation. Also @retep998 might have some valuable insights on the error codes.

r? @steveklabnik
cc @retep998
cc @Mark-Simulacrum
…ulacrum

Do not send a notification for P-high stable regressions

This is kind of a hack to only match nightly and beta regressions, but not stable regressions. See my tests [on the playground](https://play.rust-lang.org/?version=stable&mode=debug&edition=2018&gist=6ff8a809162118aa2951f2ff12400067).

r? @spastorino cc @Mark-Simulacrum
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.

8 participants