Re: [PATCH v3 00/10] upload-pack: reduce lock contention when writing packfile data

Am 10.03.26 um 14:24 schrieb Patrick Steinhardt:

I looked at these two patches.

At first, I was thrown off by the early exits when bytes_written == 0. But I convinced myself that it makes sense:

- In the emulation, if the underlying write(2) does return 0, it makes
sense to signal a partial successful write to the caller of writev. It's
the best we can do when earlier blocks have already been written. A
return value of 0 is strange and may be an indication of future
problems, so let's just stop the work we are doing and let the upper
layers deal with the problem.

- In writev_in_full we need to treat the case specially in order to
guarantee forward progress.  Treating it as ENOSPC is plucked out of
thin air, I guess, but not unreasonable.

In conclusion, these patches look good.

-- Hannes