Re: [appveyor] Really slow turnover

From: Vinnie Falco (vinnie.falco_at_[hidden])
Date: 2019-12-19 02:47:44

On Wed, Dec 18, 2019 at 4:20 PM Edward Diener via Boost
<boost_at_[hidden]> wrote:
...Attempting to run CI using Appveyor literally
> has jobs queued pending a run for days and days. Surely there must be
> something better in this day and age than this ridiculously slow
> internet service. Is it 1979 again ?

Yeah I feel you there. Beast is always right on the edge of timing out
or running out of memory on Travis. Beast could be structured into 4
separate repositories:

boost/beast/core

boost/beast/http

boost/beast/websocket

boost/beast/zlib

Each repository would have its own set of tests, documentation, and
includes, but they would share the boost/beast directory prefix. This
would complicate the tooling and there are questions about how this
would need to be presented in terms of a formal review. I would like
to eventually split Beast up, but there is resistance to this idea
from up the chain of command.

For my new libraries I focus on avoiding this problem in the first
place, by making them focus narrowly on functionality, making them be
smaller, and being mindful of compilation resources (for example, by
avoiding templates in the design, where it is practical to do so). I
am also exploring more exotic forms of type-erasure to reduce function
template instantiation.

Thanks

Thanks