nexus: fix GroupId::operator!= typo and UAF in pack_iodelay

[taskx6004]

Summary

• archdefs.h: GroupId::operator!= uses y == other.y instead of y != other.y in the last term, making the operator always return true when only y differs. This is a copy-paste error from operator==.
• pack.cc: base_iodelay_rules[id_DELAYA] = base_iodelay_rules[id_DELAYB] is undefined behavior — operator[] on the LHS may rehash the dict, invalidating the reference returned by operator[] on the RHS. Fixed by copying the value to a local before inserting.

Both bugs were found while working on LIFCL-33U bringup but are device-independent.

Test plan

• Verify nexus builds with no warnings
• Run existing nexus test designs

🤖 Generated with Claude Code

[gatecat]

I think these are reasonable changes but I'd need to check the rehashing part (also no need for the std::move in the usual nextpnr style for non-perf-critical sections).

However after the amount of upset it caused last time I'd prefer it if you resubmitted this without Claude.

As for the various LIFCL-33U PRs, this work is currently ongoing in gatecat/prjoxide#49 along with bigger prjoxide changes and more extensive testing/change handling so I won't be merging that.

[taskx6004]

fair enough. If need brute force with something, I have the board and time for help.

…

On Sun, Mar 8, 2026 at 5:36 AM myrtle ***@***.***> wrote: *gatecat* left a comment (YosysHQ/nextpnr#1661) <#1661 (comment)> I think these are reasonable changes but I'd need to check the rehashing part (also no need for the std::move in the usual nextpnr style for non-perf-critical sections). However after the amount of upset it caused last time I'd prefer it if you resubmitted this without Claude. As for the various LIFCL-33U PRs, this work is currently ongoing in gatecat/prjoxide#49 <gatecat/prjoxide#49> along with bigger prjoxide changes and more extensive testing/change handling so I won't be merging that. — Reply to this email directly, view it on GitHub <#1661 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB3EWQAMQDJ33QFVKJXYNS34PU5K3AVCNFSM6AAAAACWK3FNX2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DAMJYG4ZTAMBYHA> . You are receiving this because you authored the thread.Message ID: ***@***.***>