|
Post by elmer on May 31, 2022 6:18:10 GMT
We're going to have to debug tgemu to find out what broke. That may take some time. Or maybe not so long. The tgemu problem on linux is fixed, and "make test" works properly again ... but still only on linux, the way it's been since Uli added the testsuite.
|
|
gator
Deep Blooper
Posts: 24
|
Post by gator on May 31, 2022 21:03:36 GMT
cool
Can we move the platform test in the mk script instead of the makefile? I think running huc on Win and Mac (even if not calling tgemu after) still has some value
|
|
|
Post by elmer on Jun 1, 2022 1:46:57 GMT
Can we move the platform test in the mk script instead of the makefile? I think running huc on Win and Mac (even if not calling tgemu after) still has some value I *really* don't want the testsuite to be half-assed but yet give the impression that it's doing a full job. Anyway, I've just got confirmation that the Mac now builds cleanly with only the default XCode version of gmake, and so we should no longer need the "brew install gmake" in the github actions. Also, the full testsuite (and build) runs fine on the Mac, both x86_64 and M1 ARM. The build and testsuite also runs correctly on FreeBSD ... if the gmake package is installed. It won't build on FreeBSD with the default "bmake", because the syntax for conditionals is too different. At this point, there's only Windows to fix (or ignore), plus fixing the "make test" in the main makefile to operate on non-BASH shells.
|
|
|
Post by elmer on Jun 1, 2022 13:12:55 GMT
FYI, gator : I've fixed the Windows build of tgemu, and "make test" now runs totally fine on my computer ... but it is still producing a lot of errors on github's automated CI, almost as if it was still running tgemu from the previous check-in. Also, note that the failures in Windows "make test" run do not cause the CI to flag an error, which makes the automated test a bit pointless at the moment.
|
|
gator
Deep Blooper
Posts: 24
|
Post by gator on Jun 1, 2022 13:18:12 GMT
That good news, as it's gonna simplify the setup I don't think github provides FreeBSD runners, so unless we provide our own runners, we can't test it automatically I'll be spending the next few days in airplanes and airports, with limited access to technology, I'll look at it next week
|
|
gator
Deep Blooper
Posts: 24
|
Post by gator on Jun 1, 2022 16:14:07 GMT
Looks like you fixed the tgemu artifact download on windows. and the false green I refreshed my repo with your latest changes, and I can confirm windows tests are properly flagged as red when they fail
|
|
|
Post by elmer on Jun 1, 2022 20:22:23 GMT
I refreshed my repo with your latest changes, and I can confirm windows tests are properly flagged as red when they fail Yay! "make check" has been changed to use openssl, so I've removed all of the extra package installs (gmake, md5sum, dos2unix) from the CI actions. <EDIT> And "make", "make check" and "make test" all run fine on FreeBSD now that some BASH-specific syntax was removed from the shell scripts. This is all looking pretty clean and universal at this point!
|
|
gator
Deep Blooper
Posts: 24
|
Post by gator on Jun 7, 2022 0:02:12 GMT
Finally back online Nice to see this is coming along nicely I haven't seen much movement on the publishing flakiness, so I'll see about retrying ourselves in case of failures Also, I'd love to see tests being run on freebsd It seems to be possible to do using virtualization on mac (since mac is a close cousin) more stuff to play with
BTW, is the generated windows package good enough to be used in pinned thread ?
|
|
|
Post by elmer on Jun 7, 2022 19:06:53 GMT
BTW, is the generated windows package good enough to be used in pinned thread ? Yep, done!
|
|