2026-09-06

Not written in the café because that's closed for a vacation until 2026-09-19.

• Treat alias instances that point at the ACT_END Alias Command type similarly to stopped alias instances.
• Optimise ACT_END checks in Control-C handling, add to bx instead of loading si to address the acType field using bx + si addressing.
• Update the manual section eldcomlevels with the new EXIT type for alias instances that pointed to the ACT_END alias command, and mention initially that the comlevels handler is used for Control-C handling as well.
• Update Control-C handling protocol comments in the manual section eldcomlevels, and sync the manual's protocol comments to most of the callers. Particularly the debug.asm Control-C handling callers had some wrong comments, others just updated to be more complete.
• Add to eldcomlevels manual section and the protocol comments in the source texts two remarks, about the command line RC buffer execution not sending a comlevels callout and the fact that RE buffer execution isn't properly set up if the RE buffer content equals "@R".
• In alias.asm add references to the manual to the comlevels handler callers.
• Add acceptance tests for installing several alias.eld instances in a row.
• In Extension for lDebug loader variant used by the library ELD, allow to load the payload's code image into the data area temporarily if the code section is too full to hold both the library's loader part, the image, and the relocator. This allows the prior test to succeed in loading 3 alias.eld instances from the uncompressed extlib.eld library.
• Optimise setting eldiStartCode field in the ELD library loader. Use the es segment override prefix only if loaded into the code section, otherwise jump to such an entry that no segment override is used.
• Disable an expensive houdini breakpoint. This at some point needed a bunch of NOP instructions to succeed assembling the extpak.xld optimised packed ELD library. When this failed to assemble, the assembler would give an error "extpak.asm: error: unable to find valid values for all labels after 1005 passes; stalled for 1001, giving up."
• In the ELD loader, count the relocator space as being available for code allocation if the allocation is larger than the code image. (Is there any ELD where this is true? Answer: Does not appear to be, all extlib.eld entries have exactly the same code image and code alloc sizes.)
• Optimise some branches in the ELD library and loader code.
• Add manual section eldexelib-loader, to describe the workings of the ELD library loader a bit.
• Fix acceptance tests to actually run multi2 and multi3 tests. These were disabled by poor copy paste work on 2026-08-30 (multi2, multi3), so that the tests that were supposed to run the multi2 and multi3 test cases would actually just run the multi1 test case again.
• Fix multi3 test to actually succeed if the malias Extension for lDebug is used. Two checks had to be adjusted because the multi-alias compatibility for single-aliases needs slightly more alias buffer space than a true single-alias.
• Add two new multi-alias test cases.
• In malias_multi3 test check for nonzero ERC one more time in one of the parts specifically added for the malias Extension for lDebug. The prior check shouldn't succeed at that point.
• Document in the manual the changes needed for test_ext_malias_multi3.
• Document in the manual the multi4 and multi5 tests.
• Document in the manual the test_ext_alias_several_* tests.