On 11/12/2022 9:05 AM, Brennan Ashton wrote:
I can probably help revive my past effort to run some native builds in CI to at least prevent it from getting fully broken. One of the blockers last time, was just how slow it was and the limited number of Windows workers.
Did we ever really have a native windows build in CI? I don't think that the native build has been working since NuttX moved to the ASF incubator. Are you sure this wasn't a Cygwin build? The Cygwin build is not a "native" build since it still depends on the POSIX environment provided by Cygwin. So it is really just like the Linux build but slower. Cygwin is slow due primarily to the emulation of fork() which does not exist in Windows. But would not think that a truly native build would depend on the emulated fork() and should not be so slow.
Last time I measured, I used a script with a loop that forked some trivial process as fast as possible. The emulated fork() limited performance to about 7 forks/sec under Cygwin.
Some "native" Windows tools are still based on Cygwin (like MinGW which is usually bundled with MSYS/MSYS2) are still based on Cygwin and probably still have some fork-related performance issues. If someone really cared about the native build performance, a Visual C++ port might be work considering.
There was alsoIAR/uVisionIDEsupport in the repository for STM32. All of the IAR/uVision helper tools are still present (I think), but the IAR board support was removed https://bitbucket.org/patacongo/obsoleted/src/60ec01456d8b09f5b813a5fd8865cdbd5a0ccc20/ChangeLog#lines-357
