R devel - Jun 2022 - Stack imbalance warnings while using the grid package, in R 4.2

On 6/29/22 16:53, Panagiotis Skintzos wrote:
> Hello to all.
>
> In ggiraph package, we are getting some warnings about "Stack 
> imbalance in NextMethod" when we subclass a primitive grob and 
> override the widthDetails function. This started to occur from R 
> version 4.2 onward, in various platforms (linux, iOS, Windows).
>
> You can see the bug report, where I've added a minimal reprex:
> https://github.com/davidgohel/ggiraph/issues/224
>
> I've identified the issue in grid package. It occurs when calling 
> grid.Call.graphics with C_moveTo, C_lineTo, C_lines, C_segments, 
> C_arrows, C_polygon, C_path, C_circle, C_rect and C_raster at least. 
> The relevant functions in grid.c have mismatches in protect/unprotect 
> calls.
>
>
> As I see in the code diff for grid.c, between R 4.1 and R 4.2, one 
> extra PROTECT call was added to the above functions (L_*) but the 
> UNPROTECT call at the end of the functions was not updated.
> In other functions, like gridPoints for example, the UNPROTECT call 
> has been updated, so the issue does not occur when overriding the 
> pointsGrob.
>
> I can try to submit a patch if you wish, but perhaps there are other 
> functions elsewhere with the same issue that need to be checked.
Thanks, yes, could you please send a patch for R-devel? So that it is 
clear which PROTECTions you have in mind, and that they haven't been 
fixed already in the meantime in R-devel (note that there have been 
changes recently).

It doesn't matter if you don't fix them all.
> Hopefully this can be resolved for the next version.
Sure, PROTECT errors have very high priority.

Thanks
Tomas
>
>
> Thank you,
>
> Panagiotis
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel

On 29/6/22 17:18, Tomas Kalibera wrote:
>
> Thanks, yes, could you please send a patch for R-devel? So that it is 
> clear which PROTECTions you have in mind, and that they haven't been 
> fixed already in the meantime in R-devel (note that there have been 
> changes recently).
>
> It doesn't matter if you don't fix them all.
>
>> Hopefully this can be resolved for the next version.
>
> Sure, PROTECT errors have very high priority.
Fantastic, thanks.


Indeed most of them have been fixed in trunk.

I attach a small patch (against revision 82535) that fixes the issue for 
the remaining functions (L_moveTo, L_lineTo, L_lines).

With that applied, all tests in ggiraph pass without any warnings.


Regards,

Panagiotis



-------------- next part --------------
A non-text attachment was scrubbed...
Name: stack_imbalance.patch
Type: text/x-patch
Size: 600 bytes
Desc: not available
URL:
<https://stat.ethz.ch/pipermail/r-devel/attachments/20220629/749a9c23/attachment.bin>