Issues

Fragile quiet-mode comparison. GLib log levels are bitmasks, not an ordered scale — higher numeric values mean less severe:

G_LOG_LEVEL_ERROR    = 4
G_LOG_LEVEL_WARNING  = 16
G_LOG_LEVEL_INFO     = 64   ← what NOTE uses
G_LOG_LEVEL_DEBUG    = 128

The condition:

if (quietMode && level >= G_LOG_LEVEL_INFO)
    goto handle_fatal;

happens to suppress INFO and DEBUG correctly by accident, but reads as if INFO is "high severity". It also silently suppresses G_LOG_LEVEL_DEBUG without any documentation. The correct idiom is an explicit bitmask:

if (quietMode && (level & (G_LOG_LEVEL_INFO | G_LOG_LEVEL_DEBUG)))
    goto handle_fatal;

goto violates codebase style. main.c and the rest of the codebase don't use goto. The handle_fatal: label jump can be replaced with a plain if/else without any loss of clarity.

One exit(1) path may not close the log file. The fclose(logFile) calls are added before exit(0) and at GTK cleanup, but the exit(1) at "A valid file was not loaded" (after argument parsing, after the log file is already opened) appears to be missed. On most systems exit() flushes stdio anyway, but it is an inconsistency.

"message" prefix may confuse users. GERB_MESSAGE() uses G_LOG_LEVEL_MESSAGE which the PR labels "message:" on stderr. Most Unix CLI tools use "info:" for that severity. Minor naming issue but worth considering.

--quiet suppresses notes during pass 1 of getopt only from pass 2 onward. Any messages emitted during pass 1 of the two-pass getopt loop won't be suppressed because quietMode is set in pass 2. In practice almost no messages are emitted during option parsing, so this is unlikely to matter.

Updating command line switches should update the man page as well. I'm not sure if is up to date, but from now I think it is good idea to keep that updated as well.

Thanks for the thorough review. All six points addressed:

1. Fragile quiet-mode comparison — Replaced level >= G_LOG_LEVEL_INFO with explicit bitmask (level & (G_LOG_LEVEL_INFO | G_LOG_LEVEL_DEBUG)).

2. goto violates codebase style — Replaced goto/label with an if/else block. The quiet-mode check now wraps the print logic in a negated condition, and the fatal handler runs unconditionally after.

3. Missed exit(1) log file closure — Added if (logFile) fclose(logFile) before both exit(1) paths at "A valid file was not loaded" (single-file export check and default export switch case).

4. "message" prefix — Changed G_LOG_LEVEL_MESSAGE prefix from "message" to "info".

5. --quiet only from pass 2 — Added case 'q' to the pass 1 getopt loop so quietMode is set before any messages can be emitted during pass 2 option parsing.

6. Man page — Added -q|--quiet entry and updated -l|--log description to reflect the new behavior (writes prefixed messages to file in addition to stderr).

Force-pushed with --force-with-lease.

One concern: I'm uneasy about the naming where G_LOG_LEVEL_MESSAGE maps to "info" and G_LOG_LEVEL_INFO maps to "note". It feels inverted — "message" becoming the more informational label and "info" being demoted to "note" reads backwards. Open to suggestions on better naming here.

I agree with you. You can swap them around if you want to. I don't know if that will cause any problems in other part of the code. If it is too many changes in other places of the code maybe we should leave it at the moment and drop an issue for future reference?

It would also be great if this could be documented somewhere. Currently I am a bit overwhelmed by all PRs coming in (I am positive, don't get me wrong). But keep this in mind for the future. ;)

@spe-ciellt No urgency at all on any of our PRs — take your time. We really appreciate the decades of work you've put into gerbv, and the credit and attribution in the project. It's great to see a release happening after so long!

On the naming swap: I checked the scope — GERB_MESSAGE appears ~45 times across 10 files, GERB_NOTE is only in gerb_stats.c. So swapping them would be a mechanical rename but it touches a lot of call sites. I think the cleanest approach is to do the swap (rename GERB_MESSAGE → maps to "note:", GERB_NOTE → maps to "info:") in a separate follow-up PR so this one can land as-is. I'll drop an issue to track it and document the log level mapping there.