Credit for this goes to Stephan Beal (the eminent repo archaeologist) plus whichever "anonymous" posted this ticket, way back in 2010! ... (check-in: 6f94da9b user: wyoung tags: inskinerator-modern-backport)

Inskinerator Modern uses that rule to shift all the ancillary colors when you shift one of the primaries, saving the user from having to specify each derived color name manually. Fossil's hard-coded skins don't have that same motivation, and as hard-coded defaults, the old "office highlighter" colors were fine, but once you get down to specific instances, we don't want to use arbitrary color, we want them to be analogous or complementary to the main accent color. ... (check-in: 285defac user: wyoung tags: inskinerator-modern-backport)

This change increases the body font size for desktop use without affecting mobile. With middle-aged eyes, the old default puts too much text on a line, decreasing readability due to long line lengths, even with browsers no wider than a sheet of office paper. For our definition of "mobile," line lengths are naturally limited to reasonable lengths, and in any case, compromise is the nature of that game. ... (check-in: 2fdfb834 user: wyoung tags: inskinerator-modern-backport)

Based on discussions in forum thread 58bef43842be6fc3. ... (check-in: f9686d02 user: preben tags: hash-if-mtime-is-current)

csrf started to appear in a query due to csrf-defense-enhancement: form_begin() unconditionally generates a hidden input with csrf token since [88a402fe2a5641]. This check-in addapts /whistory page to the new behaviour. ... (check-in: abdf2805 user: george tags: trunk)

Invoked with tclsh, the script reads either a file argument or stdin and writes to stdout. ... (check-in: b3f89945 user: preben tags: testing-improvements)

The test changes now check the reverse condition: traversing symlinks should not work.

To carry out the tests, ensure allow-symlinks is set in the local repo. ... (check-in: d30cd073 user: preben tags: test-fixes-2.24)

When enumerating mxr triples, looping over multiple triples may exceed mxr. ... (check-in: 3c9472e5 user: preben tags: long-diff-fix)

OpenBSD compilation warnings (sprintf(), strcpy() and rand()) goes to stdout and ./configure then uses the warnings as part of the path to execute. ... (Closed-Leaf check-in: c8a1cfb6 user: preben tags: openbsd-jimsh0-warnings)

• FOO can now point to both a flat directory as before or to a directory one level up with lib/ and include/ subdirs, as zlib installs to when configured with a --prefix, fixing an issue reported on the forum.
• The old method would check the default paths before trying what you gave it, preventing local overrides when you know the platform version isn't what you want and you've built from source, then pointed Fossil at the replacement. It isn't always safe or advisable to remove the platform version in this case merely to get Fossil's configuration script to move on and obey your explicit directive.
• The old method checked for the header and the library in a single step, resulting in unhelpful configure output, requiring that one dig through config.log to figure out what happened.

As discussed with Stephan in /chat. File objects in the renamed case possess a 'priorName' property in addition to the 'name' reported in all cases. ... (check-in: f8d6e8c1 user: mark tags: trunk)

As reported by larrybr on the forum: 9e9778f2e6, 'fossil json status' reports renamed _and_ edited files as only renamed. This change reports such cases as both edited and renamed, and displays the filename prior to being renamed as well as the renamed path. We now also update the vfile table before generating the status report to ensure the current checkout state is reported. ... (check-in: 771e592b user: mark tags: trunk)

In `fossil {changes,status}` and the commit description shown in the editor with `fossil commit`, show the original and renamed path annotation if the file has been renamed and edited. Prior to this, the file was only reported as edited, which can be confusing. Reported by James Cook 5a4c530e6b. Discussed with stephan@ and danield@, plus some forum members. While here, make sure we honour the relative-paths setting/option when displaying the original pathname in the renamed case. ... (check-in: ca9d0ddf user: mark tags: trunk)

This fixes an issue reported by James Cook in d162b6e2dc. Previously, we assumed the fx_git table does not exist when creating a new import, thus the previously untested use case of importing multiple new (read: not incremental) git repos into the same fossil repo failed at this assumption. While here, document the persistent nature of attributions in the import cmd help text; previously, this was only in the online docs: /doc/trunk/www/inout.wiki ... (check-in: 47fb3ddc user: mark tags: trunk)

Tested on OpenBSD 7.3-current and macOS Ventura 13.4 by me, and also on some Linux machines by stephan@

This automatically generates a compile_commands.json file on machines that support the -MJ clang compiler option, which can be consumed by language server clients. ... (check-in: 91f0f00f user: mark tags: trunk)

Reported by James Cook in 2b857ded1b: renamed files were shown as a deletion and an addition. This was because the optional 3rd and 4th field in the manifest's F card to indicate renames were omitted, despite having parsed these data from the git fast-export marks. ... (check-in: 7ab1e118 user: mark tags: trunk)

As per James Cook's report in 2b857ded1b, the /info page provided no hint that a renamed and modified file had been renamed, only modified. This slightly changes the annotation to make it clear the file was also renamed. Discussed with larrybr and stephan in /chat. Slight tweak on stephan's suggested phrasing.

ok stephan@ ... (check-in: 495a4d71 user: mark tags: trunk)