These methods are implementations of the public C API and their
specification is identical:
https://cs.chromium.org/chromium/src/mojo/public/c/system/

e.g. WriteData is an implementation of MojoWriteData:
https://cs.chromium.org/chromium/src/mojo/public/c/system/data_pipe.h?rcl=3fa650bf8b19336ebb3e6748296bb060df592442&l=171

There is no reason for anyone to use the internal Dispatcher APIs directly.

Reviewing code which used DataPipeConsumerDispatcher/ DataPipeProducerDispatcher, that's not at all clear.

Again, nobody should be using these APIs directly. They are internal
details of Mojo. Can you point me at the CL?

Sorry, I was digging into the top layers of Mojo code, since that information is not included in the ScopedDataPipeConsumerHandle docs, or the MojoWriteData docs...and ended at a location with no docs, which seemed rather surprising.

I've send a lot of code assume that all data was written, and wasn't sure if was correct or not.

Hmm. https://cs.chromium.org/chromium/src/mojo/public/c/system/data_pipe.h?rcl=6fcc593c5778cabefd4ea1eeaaacedb14dc8eba0&l=151 documents that on success, *num_bytes will contain the number of bytes actually written.

I suppose we should be more explicit in pointing out that this can be less than the input value.

Hrm...  In that case, there seems to be a *lot* of broken consumer code.

https://cs.chromium.org/chromium/src/chrome/browser/extensions/chrome_url_request_util.cc?type=cs&l=230
https://cs.chromium.org/chromium/src/chrome/browser/plugins/plugin_response_interceptor_url_loader_throttle.cc?type=cs&l=47
https://cs.chromium.org/chromium/src/content/browser/devtools/devtools_url_loader_interceptor.cc?type=cs&l=904

Those are literally the first 3 non-test hits.  The third is at least kind enough to DCHECK on error, or if not all data is written, I guess?

It is safe to assume that if you created the pipe with a certain capacity,
and the pipe is still empty, a WriteData call which does not exceed that
capacity will always succeed. This is indeed not called out explicitly in
the documentation though.

Will always succeed *and all bytes will be written*, that is.