add function read_blocking/write_blocking and alike (fixes also a dep…

[craff]

I wanted to test with these as my project Simple_httpd uses non blocking sockets. I observed a 2-3% amelioration in very stressing condition (100 downloads simultaneous, using only 6 domains/cores)

So it may be worth a PR?

(fixes also a deprecation in a tests)

If you wish to accept this PR, I will add the necessary tests.

[craff]

My latest commit fixes #102 : a bug in Ssl.flush with non blocking socket.
The return case of -1 which means retry should be handled.

[craff]

I did some timing. On medium size (50Mb) size transfer, no releasing the lock is new to a factor 2 gain, because it saves releasing the lock. On simple_httpd, using domains, we go from
37 request per seconds to 57. This is for the whole process. So just timing Ssl.write and Ssl.read should indeed give a factor around 2.

[craff]

Sorry, but I am about to release simple_httpd. I do not want to fork ocaml-ssl (I think this is rude and bad for the comunity). In the meantime, I add ocaml-ssl with this PR and version updated to 0.6 as a sub module. If we converge to an accepted PR, I will remove this submodule.

[craff]

I used rebase to cleanup the commit list for easier review.

[toots]

This PR looks fine to me on principle and first pass and, if the performances you mention are real, would be warranted. We passed control of this module over to @anmonteiro and co. so I believe it would be up to them to accept the patch if they want to. Consider this as my emeritus maintainer approval

[anmonteiro]

This PR has seemingly grown in scope since it was first opened.

@craff would you mind reverting the opam file change, and submitting the deprecations in a separate PR?

[craff]

Antonio Nuno Monteiro writes:

This PR has seemingly grown in scope since it was first opened. @craff would you mind reverting the opam file change, and submitting the deprecations in a separate PR?

Done: PR 106 and PR 107 (for deprecation). PR107 is base on PR106, there was conflict to rebase on master and I think it is ok like that, I will rebase PR107 on PR106 to follow change asker by reviewver. There is a build error or a deprecated warning in a test related to Thread.Exit. One should decide if we keep the warning or if we ignore older ocaml version ? Cheers, Christophe

…

-- Christophe Raffalli tél: +689 87 23 11 48 web: http://raffalli.eu Ce mail est signé avec pgp (Pièce jointe signature.asc, clef sur https://raffalli.eu/pgp) This mail is signed pgp (Attachment signature.asc, clef on https://raffalli.eu/pgp)

[craff]

I close this PR as it is replaced by the two others.