a22b7508acMoved the freeMagic() of the labelList created for isolated substrate regions into extHardFreeAll(), where it is easy to detect if it has already been freed along with the rest of the labeled regions, to avoid a double free() call.
master
8.3.676
magic-8.3
R. Timothy Edwards
2026-07-17 13:39:51 -0400
42b346e318Extended the function of the "findlabel" command; previously, the documentation implied that "findlabel -glob <pattern> <n>" is a meaningful set of command options, but in fact was not implemented and did not raise an error for wrong number of arguments. Instead, this has now been implemented to mean find (and set the box position to) the <n>th label containing <pattern>.
R. Timothy Edwards
2026-07-17 12:07:25 -0400
cc4da9a05fCorrected a subtle error in substrate extraction: In extHierSubstrate(), extFindNodes() is called to find just the substrate node, then ExtLabelRegions() is called to label the substrate node if such a label exists. The ExtLabelRegions() routine has a section at line 344 in which if it finds a label on the substrate plane that is over space, then it attaches the label to the default substrate region. However, if the label is inside an isolated substrate region, then it passes this check, which never actually detects whether or not the label is over space. Doing so is actually simple, since the preceding code detected connecting tiles under the label, and that fact just needed to be carried forward and checked. The upshot of the error was that if an isolated substrate region like a pwell inside deep nwell was *labeled*, then it would mysteriously get shorted to the global substrate, which is clearly wrong. Thanks to Mark Martin for providing the failing example (and apologies to Mark Martin for spending time trying to debug a very obscure problem).
8.3.675
R. Timothy Edwards
2026-07-11 17:09:26 -0400
17ac06a24aFixed the handling of printed units when "units grid" is used. The original implementation (which was added recently) used an inverted output scale and multiplied by it instead of dividing by it, resulting in incorrect units. However, it was not even getting to that point due to the use of uninitialized variable "crec", which should have been set from the window pointer but was missing.
8.3.674
R. Timothy Edwards
2026-07-10 19:59:05 -0400
23a1c501ffUpdated the version to go along with the merge of pull request #540 from Darryl Miles (makes some updates to the code in the oa/ directory so that it at least will compile, even if it is not usable).
8.3.673
R. Timothy Edwards
2026-07-09 10:41:44 -0400
c6896e2a74oa: apply code-review feedback — strcpy_s link fallback + argv cleanup
Darryl L. Miles
2026-07-06 16:47:35 +0000
f3f824a6bbmagicOA.c: strcpy_s declaration (FIXME remove/move this)
Darryl L. Miles
2024-10-04 20:28:13 +0000
dfa28d1cd3oa: replace strcpy() with strcpy_s()
Darryl L. Miles
2024-10-04 20:26:12 +0000
e5d6d3f429oa/fake_oa_stub (header file oaDB.h)
Darryl L. Miles
2024-10-04 20:22:03 +0000
9ae91bb4c9oa: make it build (against stub)
Darryl L. Miles
2024-10-04 20:14:01 +0000
d9e76f372b.gitignore: add .deps/
Darryl L. Miles
2024-10-04 20:10:54 +0000
fd676c5b10git rm -rf oa/.deps/*
Darryl L. Miles
2024-10-04 20:07:03 +0100
767cde0efcTwo fixes: One by Claude Fable 5 which identified the two possible degenerate cases when attempting to find resistors in triangle topologies in ResMerge, which fixes the segfault in github issue diagnostic message. The other fix is part of github issue #538, which found an error in the check for tied transistor sources and drains. However, the other part of that issue, having to do with nondeterminism in the "extresist" output, has not yet been addressed.
8.3.672
R. Timothy Edwards
2026-07-08 13:25:01 -0400
07d98a33c2Fixed a negative tile width calculation in ResMakeRes.c, as pointed out by github user "LegumeEmittingDiode". I agree with the suggested fix and have implemented it exactly as suggested in github issue #537.
R. Timothy Edwards
2026-07-08 09:51:37 -0400
d6438387adFixed a second emscripten 6.0.2 regression in the WASM build, this one at runtime rather than at compile time. emscripten 6.0.2 removed wasmBinary (along with a batch of GL/SDL members) from the default INCOMING_MODULE_JS_API list. The JS loaders in npm/examples pass Module.wasmBinary to embed the .wasm binary, and because the WASM build links with -sASSERTIONS=1, the now-unrecognized member triggers a hard runtime abort: "`Module.wasmBinary` was supplied but `wasmBinary` not included in INCOMING_MODULE_JS_API", failing every example test. Fixed by explicitly setting -sINCOMING_MODULE_JS_API in toolchains/emscripten/defs.mak to emscripten's full default list plus wasmBinary. Spelling out the whole default (rather than only the members our own loaders use) keeps external consumers of the npm package working if they supply locateFile, arguments, and similar Module options.
Intubun
2026-07-07 14:14:21 +0200
28483435f1Fixed the WASM (emscripten) build, which broke starting with emscripten 6.0.2. The failure is not in magic itself but in the TCL dependency build: emscripten 6.0.2 began shipping a <sys/epoll.h> stub in its sysroot (6.0.1 did not), so TCL's configure now detects epoll on the Linux CI host via AC_CHECK_HEADERS([sys/epoll.h]), defines NOTIFIER_EPOLL, and compiles tclEpollNotfy.c. That file also requires <sys/queue.h>, which emscripten does not provide, so the build died with "fatal error: 'sys/queue.h' file not found". Fixed by passing ac_cv_header_sys_epoll_h=no to TCL's configure in toolchains/emscripten/build-tcl-wasm.sh, which forces the select()-based notifier. That notifier is the correct choice for single-threaded WASM anyway (emscripten's epoll is only a stub), and overriding the autoconf cache variable avoids patching the read-only TCL source tree and keeps the build working across future emscripten versions.
Intubun
2026-07-07 14:14:21 +0200
718e02ec3fClaude Fable 5 implemented capacitance checking at shape corners, after much discussion about the implementation. Since it is not exactly clear what the field lines do at corners, the implementation depends largely on an unknown and presumably global constant multiplier; this has provisionally been set to 1 but needs to be determined empirically with a field equation solver. At least one rational approximation suggests a value of pi/2 instead of 1. The corner capacitance can be considered a refinement of existing parasitic capacitance extraction, and since it takes a non-trivial additional amount of computation, it is left as an option to "extract" that can be turned on when a more accurate result is preferred at the expense of a longer extraction time.
R. Timothy Edwards
2026-07-08 09:13:07 -0400
980152ee68Fixed a second emscripten 6.0.2 regression in the WASM build, this one at runtime rather than at compile time. emscripten 6.0.2 removed wasmBinary (along with a batch of GL/SDL members) from the default INCOMING_MODULE_JS_API list. The JS loaders in npm/examples pass Module.wasmBinary to embed the .wasm binary, and because the WASM build links with -sASSERTIONS=1, the now-unrecognized member triggers a hard runtime abort: "`Module.wasmBinary` was supplied but `wasmBinary` not included in INCOMING_MODULE_JS_API", failing every example test. Fixed by explicitly setting -sINCOMING_MODULE_JS_API in toolchains/emscripten/defs.mak to emscripten's full default list plus wasmBinary. Spelling out the whole default (rather than only the members our own loaders use) keeps external consumers of the npm package working if they supply locateFile, arguments, and similar Module options.
Intubun
2026-07-07 14:14:21 +0200
ac059f0918Fixed the WASM (emscripten) build, which broke starting with emscripten 6.0.2. The failure is not in magic itself but in the TCL dependency build: emscripten 6.0.2 began shipping a <sys/epoll.h> stub in its sysroot (6.0.1 did not), so TCL's configure now detects epoll on the Linux CI host via AC_CHECK_HEADERS([sys/epoll.h]), defines NOTIFIER_EPOLL, and compiles tclEpollNotfy.c. That file also requires <sys/queue.h>, which emscripten does not provide, so the build died with "fatal error: 'sys/queue.h' file not found". Fixed by passing ac_cv_header_sys_epoll_h=no to TCL's configure in toolchains/emscripten/build-tcl-wasm.sh, which forces the select()-based notifier. That notifier is the correct choice for single-threaded WASM anyway (emscripten's epoll is only a stub), and overriding the autoconf cache variable avoids patching the read-only TCL source tree and keeps the build working across future emscripten versions.
Intubun
2026-07-07 14:14:21 +0200
443045b331Fixed a known issue in which the output of "vendor" (read-only) GDS can become corrupted if the name of the cell in magic changes from the name of the structure in GDS pointed to by the GDS_FILE property, and the new name is a different length from the original name. Fixed by Claude Fable 5 and checked for all name-change scenarios for both compressed and uncompressed output.
8.3.671
R. Timothy Edwards
2026-07-06 16:06:07 -0400
36f697d0efoa: apply code-review feedback — strcpy_s link fallback + argv cleanup
Darryl L. Miles
2026-07-06 16:47:35 +0000
4b39363a3bAdding some standard files and paths used by Claude Code so that AI metadata doesn't end up in the repository.
R. Timothy Edwards
2026-07-06 11:14:19 -0400
f7d9385c87Corrected the "char" casting of SigInterruptOnSigIO, which (1) must be able to hold the value -1, and (2) defaults to "unsigned char" on some systems. Explicitly casting this as "signed char" should avoid the problem.
R. Timothy Edwards
2026-07-06 09:17:32 -0400
3a1d63b856Fix SIGSEGV in -dnull extraction/DEF read: guard Tk_RestrictEvents with TxTkConsole
Harald Pretl
2026-07-06 11:47:12 +0800
R. Timothy Edwards
Darryl Miles
Oran C
anirban
Dan Moore
Intubun
Harald Pretl
Mohamed Gaber
Gonsolo
Andreas Wendleder
Intubun
Torleif Skår
Ahmed Nematallah
orbisai0security