vlc_interrupt: don't export vlc_*_i11e iovec API
It's not used in the code and not available/implemented on Windows.
Merge request reports
Activity
changed milestone to %4.0
added Component::Input: network streams label
assigned to @robUx4
added MRStatus::Acceptable label
We do use vector I/O in some places and there are further places that would benefit from it too.
Currently that's not overlapping with interruptible code, but that's just incidental. I don't think we should be removing a potentially useful API and undoing the compleneteness of the interruptible API set.
Even in 3.0 it's not used at all. I don't think keeping code just because we have the code is good enough reason to keep it. That means more code to maintain, to verify we update APIs. It's unlikely to be used at all since it's not available on Windows. The OSes that do have readv/writev will just use it directly (only the dtv module actually does that).
I never heard of vertorised IO before so I don't know if it's better or worse. To me any IO that is not asynchronous is bad anyway.
The removed functions are not used outside of the one local call in the same file. I wouldn't call that
used
. That's why in the end the code is merged in the single caller.Edited by Steve LhommeThe function is used, as shows in this very MR.
It seems
vlc_writev()
is used exactly once with a count != 1 here (vlc_readv()
is not).I don't think we should be removing a potentially useful API and undoing the compleneteness of the interruptible API set.
The fact that it is not implemented for Windows greatly reduces its usefulness, so maybe we could remove
vlc_readv()
andvlc_writev()
entirely (the only existing call uses it to print a\n
, it's not performance-critical).(IMO, either both (normal + i11e versions) should be kept, or both should be removed.)
added MRStatus::InReview label and removed MRStatus::Acceptable label
added MRStatus::NotCompliant label and removed MRStatus::InReview label
mentioned in merge request !2174 (merged)
mentioned in merge request !2405