runVMTests on sparc (Debug build) fails in a few places. If someone familiar with the internals could point me towards places in the source code to look at to investigate the problems, that would be very helpful.
https://paste.sr.ht/~nbuwe/5bdf787b15e8319153370ad1a7f2759010490705
RelWithDbgInfo fails during creation of the world with "must cross stacks" assertion in frame::make_full_frame_on_user_stack()
-uwe
This looks to me like a Self level error, not the VM. It is trying to find the `x11Globals` global in #10, which doesn't appear to be present. It then gives an undefined selector error in #9 Is this the same image which you were using on the other platforms? Russell
On Mon, 17 Apr 2023, at 10:01 PM, Valery Ushakov wrote:
runVMTests on sparc (Debug build) fails in a few places. If someone familiar with the internals could point me towards places in the source code to look at to investigate the problems, that would be very helpful.
https://paste.sr.ht/~nbuwe/5bdf787b15e8319153370ad1a7f2759010490705
RelWithDbgInfo fails during creation of the world with "must cross stacks" assertion in frame::make_full_frame_on_user_stack()
-uwe _______________________________________________ Self-interest mailing list -- self-interest@lists.selflanguage.org To unsubscribe send an email to self-interest-leave@lists.selflanguage.org
On Wed, Apr 19, 2023 at 13:10:27 +1000, Russell Allen wrote:
It is trying to find the `x11Globals` global in #10, which doesn't appear to be present. It then gives an undefined selector error in #9 Is this the same image which you were using on the other platforms?
It was the -o none snapshot, so there's no graphics. Also there are no processes there too, I guess, hence "This is not a scheduled process...". Unfortunately -o morphic Snapshot takes an order of magnitude more time to create (emulated sparc, debug build). Not long enough to trigger JCL flashbacks, but dangerously close to it :)
A working "headless" world that can run the test suite would be very helpful.
-uwe
On Wed, Apr 19, 2023 at 13:20:12 +0300, Valery Ushakov wrote:
A working "headless" world that can run the test suite would be very helpful.
With a -o morphic snapshot (processes), and adding SignalInterface::is_uplevel_trap that sparc requires - I get a full test run it seems:
https://paste.sr.ht/~nbuwe/500c9879151260a5b7b3f597aeb874378eeec0e4
-uwe
Hi -
On Wed, 19 Apr 2023, at 8:20 PM, Valery Ushakov wrote:
On Wed, Apr 19, 2023 at 13:10:27 +1000, Russell Allen wrote:
It is trying to find the `x11Globals` global in #10, which doesn't appear to be present. It then gives an undefined selector error in #9 Is this the same image which you were using on the other platforms?
It was the -o none snapshot, so there's no graphics. Also there are no processes there too, I guess, hence "This is not a scheduled process...". Unfortunately -o morphic Snapshot takes an order of magnitude more time to create (emulated sparc, debug build). Not long enough to trigger JCL flashbacks, but dangerously close to it :)
A working "headless" world that can run the test suite would be very helpful.
I've tracked down the issue, the headless build is trying to initialise the X11Globals but of course they aren't present in the build.
I can do up a fix which causes the system to abandon attempting to load the GUI if it can't find the right graphics global object, printing out a notification on the console so the user knows what is happening.
Maybe I should then send that to you to put in your tree? Or I could put it in the `netbsd` branch and you could rebase your code on it?
- Russell
-uwe _______________________________________________ Self-interest mailing list -- self-interest@lists.selflanguage.org To unsubscribe send an email to self-interest-leave@lists.selflanguage.org
On Thu, Apr 20, 2023 at 10:30:58 +1000, Russell Allen wrote:
A working "headless" world that can run the test suite would be very helpful.
I've tracked down the issue, the headless build is trying to initialise the X11Globals but of course they aren't present in the build.
I can do up a fix which causes the system to abandon attempting to load the GUI if it can't find the right graphics global object, printing out a notification on the console so the user knows what is happening.
Maybe I should then send that to you to put in your tree? Or I could put it in the `netbsd` branch and you could rebase your code on it?
I'm prefectly fine with runnng with local uncommitted changes, so as long as the fix is available somewhere (a commit on dev or netbsd, a patch in a private mail), it should be ok.
Since the test suite seems to want to have the scheduler running, how do I do that manually, without getting all of the morphic GUI in. What the morphic snapshot reports as:
info -- Scheduler started (according to snapshotAction schedulerInitial)
-uwe
On Thu, 20 Apr 2023, at 11:51 AM, Valery Ushakov wrote:
On Thu, Apr 20, 2023 at 10:30:58 +1000, Russell Allen wrote:
Maybe I should then send that to you to put in your tree? Or I could put it in the `netbsd` branch and you could rebase your code on it?
I'm prefectly fine with runnng with local uncommitted changes, so as long as the fix is available somewhere (a commit on dev or netbsd, a patch in a private mail), it should be ok.
Since the test suite seems to want to have the scheduler running, how do I do that manually, without getting all of the morphic GUI in. What the morphic snapshot reports as:
info -- Scheduler started (according to snapshotAction schedulerInitial)
-uwe
I'll see if I can push something to the `netbsd` branch.
The scheduler should be OK to run, it doesn't depend on the GUI.
- Russell
HI -
On Thu, 20 Apr 2023, at 12:35 PM, Russell Allen wrote:
On Thu, 20 Apr 2023, at 11:51 AM, Valery Ushakov wrote:
On Thu, Apr 20, 2023 at 10:30:58 +1000, Russell Allen wrote:
Maybe I should then send that to you to put in your tree? Or I could put it in the `netbsd` branch and you could rebase your code on it?
I'm prefectly fine with runnng with local uncommitted changes, so as long as the fix is available somewhere (a commit on dev or netbsd, a patch in a private mail), it should be ok.
Since the test suite seems to want to have the scheduler running, how do I do that manually, without getting all of the morphic GUI in. What the morphic snapshot reports as:
info -- Scheduler started (according to snapshotAction schedulerInitial)
-uwe
I'll see if I can push something to the `netbsd` branch.
The scheduler should be OK to run, it doesn't depend on the GUI.
- Russell
I've pushed a small patch (2 methods changed) to the `netbsd` branch. It should solve the specific issue, and at least on my machine I can now build the Self world with `-o none` successfully.
It will automatically start the scheduler for you.
Let me know how you go, - Russell
self-interest@lists.selflanguage.org
-
Russell Allen -
Valery Ushakov