nuttx-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From GitBox <...@apache.org>
Subject [GitHub] [incubator-nuttx] btashton commented on issue #886: Cygwin sim:nsh failed on apps/builtin
Date Tue, 28 Apr 2020 04:39:14 GMT

btashton commented on issue #886:
URL: https://github.com/apache/incubator-nuttx/issues/886#issuecomment-620376183


   > > > I just about have a CI run working for Cygwin but it is ugly and really
slow. I'll put up a Draft PR tonight and we can work on getting the Windows testing in a much
better place.
   > > 
   > > I think we should be the PR checks short and sweet. Just enough to give us some
confidence that the change is not generally lethal. To me this could be primarily Linux tests
of various configurations. More exhaustive testing should be left for the nightly build.
   > > 
   > 
   > Yes, this is how macOS do by @yamt, which just build three config(arm12, sim and mips-riscv-x86-xtensa).
   > For Windows, I would suggest that:
   > 1.Build sim for all variants(Cygwin, Msys2, Native and WSL) since sim tightly couple
with OS.
   
   Just for reference it currently takes over 1.5hr to run the Cygwin sim builds.  There are
a couple optimizations I can make but I would not expect to shave more than 30%.  WSL is just
Ubuntu so I don't see any reason to duplicate that. 
   
   The Native builds should be quite a bit faster.
   
   > 2.Select one but different arch for each variant
   > 3.Enable all config which can only build on Windows(e.g. z80?)
   > 
   > > If a Windows CI is really time consuming too much time for efficent PR checks,
perhaps only a 
   > 
   > Now, we build on Linux/macOS with 18VMs, but the finish time is around from 7m to
41m. So we have enough room to optimize VMs usage(distrube the work more balance) to 16VMs
at least, then we have 4VMs for each Windows variants.
   > 
   > > nightly test would be sufficient. That would not detect breakage on the PR, but
would still catch the breakage post-merge. Which, at least to me, is and acceptable performance/thoroughness
tradeoff.
   > 
   > It's hard to deploy the nightly build to Apache Windows server, since:
   > 1.Apache Windows server is much less than Linux server
   > 2.We don't have the right to install software directly without JNFRA help
   > 
   
   I don't really see any reason we cannot do the nightly build using GitHub Actions we can
easily make the targets conditional on the job and then we don't have to redo all the CI work
and have close to full coverage. 


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



Mime
View raw message