The recent sed update in cygwin [1] changed the way sed handles
carriage returns. This meant that every .in file we processed
still contained CR, which doesn't work for .sh files.
Force LF line endings for those files (and also other shell scripts).
[1] https://cygwin.com/ml/cygwin-announce/2017-02/msg00036.html
It'd not be recognized as a shell file so it'd get CRLF line
endings on Windows which doesn't work. Also added a shebang.
Also remove get-cygwin-deps.sh, it is long outdated.
Correct instructions are on the Mono website.
cert-sync already knew how to import into the BTLS cert store,
update it to always import into both stores so we don't need
the btls-cert-sync tool anymore.
The elif for the interpreter was done after the elif
that catches (almost) all cases so it was ignored.
Simplify and condense the CI_TAGS checking so it's
easier to understand and more readable.
Get rid of WORKSPACE which is set by Jenkins, this allows one to
more easily run the run-jenkins.sh script locally e.g. on cygwin.
The only thing that needs to be passed in is CI_TAGS and label.
* [ci] Run aot-test on more platforms
Now that https://github.com/mono/mono/pull/4096 is merged, we should be able to run it on Linux too.
* [ci] Don't run aot-test on Windows
It's broken there.
* [ci] Add aot-test step like on Wrench
* [tests] Fix mktemp argument to work on Linux
* [ci] Only run aot-test step on OSX
It seems to crash on linux with the boehm configuration.
The babysitter script used for CI results was previously able to run
either on Windows or standard UNIX. However, the builders run Cygwin,
which turns out to need special support. To make Cygwin work two
changes were made:
1. Paths which are exported to Mono get filtered through `cygpath`.
Paths received back from Mono are filtered in reverse. The script now
tracks two sets of path strings.
2. Under certain circumstances in Cygwin, pgkill() attempts to close a
process group and finds it does not exist. My assumption is that this
is because it has already been closed. I dealt with this by simply
catching the failure and printing a warning.
The additional profiles like monotouch, monodroid etc. are normal
managed code anyway and are the same no matter where they're built
so we can save a lot of CI time by not building them e.g. on slower
ARM bots.
We decided it's a better name since we're using the profile to
test FullAOT scenarios on desktop (as opposed to the mobile profiles).
While doing the rename I saw that a lot of the #ifdefs were using
MOBILE_STATIC even though the code should actually be included/excluded
in all FullAOT scenarios, e.g. System.Reflection.Emit. I replaced those
cases with FULL_AOT_RUNTIME instead.
For cases where code is specific to this profile, I added a new
AOT_ONLY_DESKTOP define.
Cleaned up Mono_Posix_FromSeekFlags() in support/map.c. It didn't properly
handle the case when L_INCR, L_SET, L_XTND aren't available, like on Windows.
Added Mono_Posix_Stdlib_GetLastError() to support/errno.c and changed
Stdlib.GetlastError() use it. Without this Stdlib.GetlastError() wouldn't take
errno in the C runtime used by libMonoPosixHelper.dll into account.
On Windows the Syscall class failed to initialize due to missing functions,
e.g. Mono_Posix_Syscall_get_at_fdcwd(), in libMonoPosixHelper.dll. Those
readonly static fields which were initialized at class initialization have now
been converted to static readonly properties so that the Syscall class can be
initialized despite those functions not being defined.
Cleaned up temp file handling in StdioFileStreamTest to avoid failures due to
sharing violations on Windows.
Fixed a bug in the StdioFileStreamTest.Write() test method when reading back
the batch of bytes written in the second part of the test.
Enabled the Mono.Posix test suite to run on PR builds. This test suite should
from now on be green.