[JENKINS-37575] Keep track of how much content has been copied by writeLog even if the callable is interrupted #74
Conversation
…teLog even if the callable is interrupted.
| raf.seek(lastLocation); | ||
| long toRead = len - lastLocation; | ||
| if (toRead > Integer.MAX_VALUE) { // >2Gb of output at once is unlikely | ||
| throw new IOException("large reads not yet implemented"); | ||
| } | ||
| // TODO is this efficient for large amounts of output? Would it be better to stream data, or return a byte[] from the callable? |
There was a problem hiding this comment.
Because the speculation is obsolete—the PR is picking another approach entirely.
| raf.seek(lastLocation); | ||
| long toRead = len - lastLocation; | ||
| if (toRead > Integer.MAX_VALUE) { // >2Gb of output at once is unlikely | ||
| throw new IOException("large reads not yet implemented"); | ||
| } | ||
| // TODO is this efficient for large amounts of output? Would it be better to stream data, or return a byte[] from the callable? | ||
| byte[] buf = new byte[(int) toRead]; |
There was a problem hiding this comment.
Still not a fan of the fact that we chose to do a complete buffer read (since potentially we could have a very large block of data). I'm itching to change this once these logging PRs are all integrated.
There was a problem hiding this comment.
#60 (when active) renders this dead code, replaced by something that (potentially) streams content.
There was a problem hiding this comment.
@jglick What happens if encoding changes alter the number of bytes needed to represent the character content (likely to happen in the interactions with the UTF-8 encoding PRs I think)? It looks like we're counting at the output level, which would be after character set transcoding, but seeking by the same amount in the log file itself (original encoding). Or am I misunderstanding how that interaction will work?
|
|
|
@jglick What is the next step for this PR? |
|
@recampbell I am awaiting reviews. |
|
I needed to double check yet another interaction with #61 to make sure this wouldn't need amendment with the changes there but I think we're fine now. |
…nly what Handler.output actually received, in case of some exception halfway followed by a fresh agent connection.
See JENKINS-37575 where there is information to reproduce. This is dead code as of #60 + jenkinsci/workflow-durable-task-step-plugin#63 but may as well fix the bug so long as the old code path can be run at least with a kill switch.