STM32CubeIDE instability with MX

[PGump.1]

Hi all,

If I recall correctly, CubeIDE V1.8 was very unstable when using the integrated MX, with many of the control components failing to redraw correctly, or not at all. Sometimes you could trick it into drawing by minimising and restoring the form. It was all very intermittent and problematic.

I was very happy see that issue fade away in subsequent versions. I think it was V1.13 that I saw it last...

However, I have seen it twice now in V1.16...!!! I really hope it's not coming back.

Yesterday I attempted to start a new project from MX - it got as far as the STM32F4 license agreement form where it looked up - all the controls on the form were greyed out, the form was system modal. I can kill the CubeIDE, but trying to kill java is like trying to kill Dracula's zombie.

Only rebooting the PC was going get it out of this. As you know, this crashes your workspace, and you have to go through all the motions to recover it.

Perhaps QC need to change their tests and measures...

Kind regards
Pedro

[Bob S]

The easy solution has always been - run CubeMX stand-alone, separately from the IDE.  Don't run it from within the IDE.

[PGump.1]

Thank for your reply.

However, I wouldn't say that dividing a single project into 2 projects is a good solution...

Kind regards
Pedro

[Bob S]

They started out as 2 separate products.  Actually, CubeMX pre-dates CubeIDE.  The mistake (i.e. bad solution) IMO was wrapping CubeMX inside the IDE.  Complexity on top of complexity for no real benefit.  They work perfectly well as separate programs.

[PGump.1]

Unfortunately, a-work-around is not a good formula for product improvement.

Complexity! Isn't that what computers are made for... 🙂

Kind regards
Pedro

[PGump.1]

This issue with MX is definitely back!

I am observing that there are circular references in the many of the form's controls. Looking at the behaviour of the device pin labels, where there are multiple controls for this for pin label value. I can see that changing the value in one control triggers an event in its associated control(s), and vise-versa... So, "around and around and around it goes, where it stops, nobody knows..."

Kind regards
Pedro

[PGump.1]

I have some more feedback on this issue.

If I run my Linux GUI as X11, MX is very slow, but usable. It goes faster if I disconnect it from the internet.

However, if I run the GUI as Wayland it is unusable -

After 45 minutes it get this far

Just the image of the device. No controls are active.

 Just the image of the device. No controls are active.

After 2 hours 30 minutes I get -

 None of the controls are active yet.

All this time Java, and its chrom'ies, are going flatout -

All my CPU cores are max'ed out. Temperature has max'ed out. The fan sounds like my laptop is going to morph into a hoover craft and take off...

I had to shut it down... The noise was interfering with some quiet time I had planned.

Kind regards
Pedro