v0.18.0

Added

• list_triggers and list_program_units name where an inherited body lives (#122). A row with bodySource: inherited now carries the same inherited pointer get_trigger / get_program_unit serve (parent module, file, name, ownerPath), at either verbosity. In a mostly subclassed form that turns one get_trigger call per trigger into a single listing.
• The "no converter" error names --convert-command / OFMCP_CONVERT_COMMAND (#122), or --compile-command / OFMCP_COMPILE_COMMAND for a .pll, beside ORACLE_HOME and the pre-converted file. The site's own command is often the easiest fix, and the error was where nobody learned about it.
• Force a re-conversion of a cached module (#121): fetch_module(force=true), and --force with --prefetch / --prefetch-all / --prefetch-type. A module whose source is unchanged was never converted again, so a result from a wrong setup — frmf2xml without the forms directory on FORMS_PATH succeeds minus every inherited object — kept being served after the setup was fixed, until its cache entry was deleted by hand. Annotations are kept. --force without a prefetch is a usage error.
• The forms directory can be given as OFMCP_FORMS_DIR, for a host that starts the server with a fixed command line and can pass a user's settings only as environment variables. --forms-dir or the positional argument still wins, and an exported variable never conflicts with either. A blank value, or a ${…} template nobody filled in, counts as not set.
• {start} in a converter command is replaced with the absolute directory the server was started in, in --convert-command and --compile-command alike, and is never appended. A relative program was already found from there, but a relative argument is looked up by the command, which runs in the converted directory; {start} names a file shipped beside the server (["java", "{start}/converter/Convert.java", "{}", "{out}"]) from one value on every OS.
• The plugin launcher runs without OFMCP_PLUGIN_DATA (plugin version 1.5.0), keeping its downloads in a per-user data directory of the OS instead of stopping. Its messages name the OFMCP_SERVER_VERSION variable beside /plugin, for hosts that have no /plugin.

Fixed

• ORACLE_HOME conversion works on a Linux Forms 12c+ install (#120). Three things stood in the way, on Forms 14.1.2 at least:
  • frmf2xml is not in $ORACLE_HOME/bin there, only as frmf2xml.sh in the Forms instance. The tools are now also looked for in $FORMS_INSTANCE/bin, and .sh (and, on Windows, .cmd) wrappers are accepted.
  • frmf2xml writes its XML next to the module there, not into its working directory, so the server found no output — and a writable forms directory got XML among its sources. It now converts a copy of the module in a private staging directory and moves the result into the cache, so the forms directory is never written to and may be read-only.
  • Started without a terminal or a display, as an MCP host may start it, frmcmp_batch failed with FRM-91500 and frmf2xml with "Failed to create new forms context". Outside Windows, the tools now run with TERM=vt220 and FORMS_API_TK_BYPASS=TRUE unless the environment already sets them.

What's Changed

• chore(deps): Bump eclipse-temurin from dbb43d8 to ecdc0ad by @dependabot[bot] in #116
• chore(deps): Bump gradle/actions/setup-gradle from 6.4.0 to 6.4.1 by @dependabot[bot] in #117
• chore(deps): Bump docker/login-action from 4.6.0 to 4.7.0 by @dependabot[bot] in #118
• feat: launch from environment variables alone (#119) by @aoreshkov in #123
• fix(convert): ORACLE_HOME conversion on a Linux Forms 12c+ install (#120) by @aoreshkov in #124
• feat: force a re-conversion of a cached module (#121) by @aoreshkov in #125
• feat: name inherited sources in list rows and the convert command in the no-converter error (#122) by @aoreshkov in #126
• chore(release): 0.18.0 by @aoreshkov in #127

Full Changelog: v0.17.0...v0.18.0