shared/opam_install.sh is a heresy

[maximedenes]

Correct me if I'm wrong, but with shared/opam_install.sh, if the network is down for 15 minutes, we'll get results without failures, but with a 15 minutes increment in the timings. That's a terrible idea, and that's probably what happend with fiat some time ago.

[ghost]

Good point.

[ghost]

Two things (at least) need to be done:

• avoid non-terminating loop in case of permanent network errors
• make sure that if transient network errors occur, the benchmarking time is not disturbed.

This is related to two problems:

• originally, when some of the opam install commands failed, the whole thing failed
• there is no direct way to separate the download action from the build and install actions

If they were separable, I could wrap the download action with a loop that would deal with transient networking problems so that we can measure the time of build independently from the time how long does it take to download a given thing. This is related to issue #1 .

[ghost]

This should be fixed now.

My very rudimentary tests have finished. No gross errors were discovered with this commit.

[ghost]

I still use the heretic opam_install.sh in those places where it is harmless, though.

[maximedenes]

I still use the heretic opam_install.sh in those places where it is harmless, though.

:) You mean to install dependencies? It would be good to limit the number of attempts in any case (like 3).

[ghost]

Ok. I'll make that change.