2023-05-14

This week Einsicht got some development, and an additional backup was set up for the pushbx blog.

The first change was just a fix to allow building Insight again, one instruction had escaped the conditional directives needed to exclude it at build time.

The other changes concern the disassembly window. It can now send commands to lDebug and read the disassembly to update its own disassembly window.

This is still an early stage. For instance, any update of the disassembly (or "command") window will discard all prior disassembly from its buffer and reread the entire window's worth of data.

Also, the symbolic displays indicating branch targets and symbols within the disassembled code are probably not handled well yet. Likewise, disassembling a 32-bit code segment won't work yet.

And this is just one single call site of Insight's disassembler replaced by Einsicht's lDebug I/O yet.

Further, hiding the input and output from automated Einsicht commands is still missing. Additionally, it is not clear how TracList should interact with Einsicht at that point.

The other development this week is a basic backup repository for the dokuwiki pages as well as pushbx blog comments.

I decided to implement an allowlist of the included content, lest I have to check all the files that could be included.

For now, the data that it operates on lives in the same repo, same named branch, and same unnamed branch as the scripting itself. This may be revisited later.

There are a number of filters run on the data to redact email addresses, URLs, and IP addresses found in the comment files. I found that these files appear to be serialised PHP data structures. I hope that the files can still be used despite my filters' changes. The filtering is done using perl scriptlets. (During these changes, I noticed a bug in diffr, the diff colour highlighting tool that I often use.)

The filtering to drop subscribers lists from the comment files assumes that no curly braces occur in quoted strings (such as the email addresses). This could be better.

The IP address detection uses a regular expression allowing hexits, dots, and colons, and checks that at least 8 of those are back to back. The correct choice would probably be 7, as there are IP v4 addresses of the form x.x.x.x where each x is a single digit, for a total of 7 codepoints.

Further, the automatic nature of the script is not yet deployed. It will only create the new commits, besides, not yet push them to the webrepos. Both of these steps need to be ran manually for now.

Finally, once pushed, the data is backed up along with all my webrepos – that is, those below the address https://hg.pushbx.org/ecm/ (does not include the SvarDOS backup). Then I can upload the backup file occasionally to the web archive's wayback machine, ensuring an independent backup that will hopefully last a while.

I intend to pick the changes to the TSR example to all my TSRs soon. I will keep /J+x as a synonym for /Jx, which will mean to "jettison" a feature.

The other upcoming thing is a review of source-available AMIS multiplexer implementations. I have been working on that several times in the last months.