check_config: validate INI datatypes (numeric / enum / duplicate) before comparison

[tzuohann]

Problem

check_config.tcl compares [JOINT_n]/[AXIS_L] limits with Tcl >/<.
Tcl compares numerically only when both operands are valid numbers; otherwise
it silently does a string comparison. So a limit line carrying stray text —
e.g. an inline # comment — gets compared as a string.

With equal limits like:

[AXIS_A]
MAX_LIMIT = 3600000                # placeholder
[JOINT_4]
MAX_LIMIT = 3600000                  # placeholder

3600000 == 3600000, yet check_config reports
[JOINT_4]MAX_LIMIT < [AXIS_A]MAX_LIMIT — because the strings (comments +
differing whitespace) compare unequal. Pass/fail flips on trailing spaces, and
the error points at the wrong thing.

Fix

Add assert_ini_datatypes, run once right after parse_ini, before any
comparison. For each [JOINT_n]/[AXIS_L] item:

• duplicate keys (multi-valued result) → error;
• numeric items must satisfy string is double -strict (ints and reals);
• enum items must match a fixed set, case-insensitively, mirroring
  LinuxCNC's caseless maps:
  • TYPE → {LINEAR ANGULAR} (IniFile::mapJointType)
  • boolean home flags → {TRUE YES 1 ON FALSE NO 0 OFF} (IniFile::convertBool)
• unknown items are left alone.

Errors name the [SECTION]ITEM and the offending value.

Behavior change

Duplicate [JOINT_]/[AXIS_] keys are now an error, replacing the warn-only
warn_for_multiple_ini_values (removed; same scope, now fatal).

Testing

Patched checker vs a working 5-joint INI and corrupted copies:

• Accepted: valid int & real, lowercase TYPE=angular, all eight boolean spellings.
• Rejected (clear message each): inline comment on a limit, abc, 1,000,
  1.2.3, TYPE=LINER, HOME_USE_INDEX=MAYBE, duplicate key.

Note

I'm not a software developer by trade, and this is my first day using LinuxCNC,
so I'd genuinely welcome feedback on the approach, the Tcl style, or the
numeric/enum item lists — happy to revise.

[grandixximo]

Thanks for digging into this, and welcome to LinuxCNC.

The inline # comment case is intentional, not a parser slip. Bertho's INI parser rewrite (April-May this year) makes # and ; comment markers only at the start of a line; after the = they are literal characters, by design. The reason is that values legitimately need them. For example the shipped qtdragon configs:

MDI_COMMAND_MACRO0 = G0 Z2;X0 Y0;Z0, Goto\nUser\nZero

The ; is a G-code command separator. If the parser stripped from #/; onward, that macro silently loses everything after Z2. So treating trailing text as a comment would break configurations that have to keep working. It is documented in docs/src/config/ini-config.adoc and pinned by tests/inifile/comments/.

Also worth knowing: the datatype, enum, and duplicate-key checks here already exist in src/emc/ini/linuxcnc_check_ini.py, the pre-run checker Bertho added in the same series. It validates TYPE, the boolean spellings, integer ranges, and duplicate keys, using the C++ parser's exposed enum functions so the accepted values stay in sync with the engine. That is probably the durable place for clearer errors, rather than reimplementing the checks in the hallib Tcl. Worth getting Bertho's read since this is his subsystem.

Given the above, I tend toward closing this, but happy to hear if I am missing something.

[BsAtHome]

The lib/hallib/check_config.tcl script has been superseded by emc/ini/linuxcnc_check_ini.py.

Any changes to the Tcl script are in vain.

[grandixximo]

@BsAtHome if check_config.tcl is fully superseded by linuxcnc_check_ini.py, why is it still in the tree? Should it be deleted, or is it kept around intentionally for now?

[BsAtHome]

There were some doubts when I wrote the new checker so I left it hanging around unused. There have been no complaints, so better remove the old Tcl script.