search

RTimothyEdwards / magic

Magic VLSI Layout Tool
Other
472 stars 99 forks source link

Regression in DRC detection #307

Closed donn closed 4 months ago

donn commented 4 months ago

Magic version 8.3.466 correctly finds DRC issues in a gds file, but 8.3.479 reports no DRC errors. Are we missing something…?

reproducible.tgz

d-m-bailey commented 4 months ago

@RTimothyEdwards 8.3.471 is the commit that stops extracting errors. 8.3.470 shows

Magic 8.3 revision 470 - Compiled on Sun May 12 06:24:50 PDT 2024.
...
DRC style is now "drc(full)"
Loading DRC CIF style.
[INFO] COUNT: 171

8.3.471

Magic 8.3 revision 471 - Compiled on Sun May 12 06:27:30 PDT 2024.
...
DRC style is now "drc(full)"
Loading DRC CIF style.
No errors found.
[INFO] COUNT: 0
RTimothyEdwards commented 4 months ago

The only thing I changed in 8.3.471 is the parsing of arguments to the "select" command. My best guess is that in fixing the parsing of some of the "select" syntax that was broken, I broke something else. Is there a critical "select" command in the DRC script?

RTimothyEdwards commented 4 months ago

If I run the script commands individually, then the issue is clearly with the "select" command since "select top cell" returns "PHY_EDGE_ROW_1_Left_14" and not "inverter". Not sure why yet---I ran other quick tests and "select top cell" appeared to be doing the right thing.

RTimothyEdwards commented 4 months ago

The problem seems to be specific to using "magic -dnull".

RTimothyEdwards commented 4 months ago

Version 8.3.482 should fix this. The "select top cell" was behaving like "select cell" because there was no specific code handling the "top" option. The change to be more systematic about handling the various arguments like "more", "less", etc., caused this issue to suddenly manifest itself. The new code simply registers the fact that "top" has been issued and selects the top level cell as it should.

I have tested this on the reproducible case and it works.

donn commented 4 months ago

Confirmed fixed. Thanks, Tim!