-
-
Notifications
You must be signed in to change notification settings - Fork 31.6k
Revert "fs: ensure readFile[Sync] reads from the beginning" #10809
New issue
Have a question about this project? # for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “#”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? # to your account
Conversation
This reverts commit 4444e73.
Which tests fail for you? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I can reproduce these failures on master
on Linux
LGTM
I’m landing this to un-break master… it’s just a revert so we shouldn’t need to wait for CI @seishun The PR description was edited here to include the relevant output of |
Landed in 66f09be |
Ok, I see why it fails. The tests call
cc @nodejs/platform-macos @nodejs/collaborators |
There might be a reasonable interpretation what that means then talking about stdin; namely, open the file and read until EOF is hit. What pretty much doesn’t make sense here is reading from a given offset. (This isn’t MacOS-specific, by the way; it should affect all POSIX systems, at least.) |
Right, in that case it would have to use the old behavior for all TTYs, pipes and sockets.
Right, if you're using |
We could probably reasonably do special-casing based on whether the fd refers to a regular file or not, yeah.
Okay, this is obviously just my personal impression, but I wouldn’t actually expect that to happen? Maybe it’s just me, but I’d definitely find it weird for the same data to be read twice from an fd. |
Are you talking about "non-regular" fds or all fds? In the latter case, the current behavior is what you want, and I'd be fine with it, but the docs need to be clarified then. |
Should we add an item to the PR template checklist about running CI for the last commit before the PR lands? (and other things like number of LGTMs, 48-hour wait, .etc) |
@joyeecheung: That likely wouldn't help much. It's just something that those of us landing PRs need to remember to check for. Thank you @addaleax for landing the revert. |
I definitely can see arguments for both perspective, but yeah, what I had in mind was having the current behaviour for all file types. |
This reverts commit 4444e73. PR-URL: nodejs#10809 Reviewed-By: Anna Henningsen <anna@addaleax.net>
This reverts commit 4444e73.
Looks like #9699 was landed without a CI Run. It breaks three tests in my
Mac OS.
Checklist
make -j4 test
(UNIX), orvcbuild test
(Windows) passesAffected core subsystem(s)
fs
cc @seishun