Code generation fails for MemoryAliasNode

[alexian79]

project.zip

Hi, I am having issue with generating machine that has MemoryAliasNode. I have attached compressed project.py. I am using QEMU 5.0.1
generation fails then checking for Type.exists, node.allias_offset is assigned to CINT object that is not string

I am able to fix this in my case by adding this check

but I am not sure that this is the right way.
BR. Oleksiy

[laerreal]

Hi! This is acceptable workaround. But real problem is in machine.py.

Previously, alias_offset could be string only (containing either integer literal or name of a macro).
Now it also can be a source.CConst's descendant.

There already are fixes for same problems in devel branch.
I also will push a solution for this issue soon (in devel).

Please note, that master branch currently frizzing because we are busy in other projects.
All bug fixes and new features now can be found in devel.
With new bugs, of course.
One day they will appear in master.
Fixes and features, probably, will appear too.

[laerreal]

Done.
Works for me with the project you provided.

[alexian79]

Thanks for the tips about devel branch, will use that.
Having a chance I would like to ask, Is there any work done on supporting qemu 6 and up, witch uses meson build? any chance devel branch was updated to use Python >= 3.9
Thanks, for replies.

[laerreal]

Hi!

I am currently developing models for 7.0.50 using devel. It should work for 6 too, probably. meson is supported.
QDT has mechanism to keep compatibility with all versions of Qemu it ever supported. However, some new features may not work with old Qemu versions (which were actual before the features are added).

Yes, we also trying to support new versions of python. While still not dropping 2.7. As soon as Py3 tends to give compatibility surprises, an idea to drop it at all and maintain Py2 only (as it will never change) looks attracting to me...

[alexian79]

Thanks for update,
I am trying to follow your advice to use devel branch, and I see your fix to reported problem, but there are issues with pulling submodules. Looks like some of them have been relocated.
I have attached log to see what is failing. I could find some of them in different places on github. Any advice on that will be appreciated.
submodule_fails.txt

[laerreal]

Try to revert this patch.
f5ebc8d
Other submodule references also should use https, as it's appear to require less rights for user.

[laerreal]

Updated devel.

[laerreal]

Now all modules use HTTPS.

[alexian79]

That did the trick, submodule update works now. Thank you.