Hello @5shekel

Thank you for you pull request. Seems like a useful thing.

Tested the branch — builds clean, no regressions (89/100 pass, same 11 failures as develop baseline). That is because Cairo versioning thing, but it builds and is accepted by the regression tests in the pipeline. The feature works correctly; layered SVG output is well-formed and opens with proper Inkscape layers.

A few things worth addressing:

1. Nested <svg> inside layer groups
The output wraps each layer in a child <svg> inside the <g>, which is non-standard for Inkscape layers. Standard practice is to put drawing primitives directly in the <g>. It works in Inkscape but may behave oddly in other SVG editors. Not a blocker for a first version.

2. Partial output on error
If exportimage_render_svg_layers_from_project() fails mid-loop, it still writes the partial SVG to disk. Should either skip writing or g_unlink() the output file on error.

3. Silent background skip
exportimage_append_svg_background() silently skips pure white and pure black — worth a comment explaining why, since it'll surprise anyone who deliberately set a white background.

Final note
When you fix this, could you merge all commits into one and then force-push it to this branch? (git push --force-with-lease)

Re-checked after the two follow-up commits (8da9640, aa22e7d). All three issues from the earlier review have been addressed:

1. Nested <svg> inside layer groups — Fixed
exportimage_extract_svg_inner_content() now extracts only what's between the Cairo <svg> opening tag and </svg>, so each Inkscape layer <g> contains drawing primitives directly. No more nested <svg>.

2. Partial output file on error — Fixed
The function builds the entire SVG in a GString in memory and only writes to disk at the end, guarded by !hadError. A failed layer mid-loop leaves no partial file on disk.

3. Undocumented background skip — Fixed
The comment in exportimage_append_svg_background() now explains that white/black are intentionally skipped to stay consistent with the non-layered SVG export path.

One minor item still open: exportimage_extract_svg_inner_content() uses strstr/strchr/g_strrstr to slice up Cairo's SVG output. A short comment noting that this relies on Cairo's stable output format would help future maintainers. Not a blocker.

Tested the updated branch against example/hellboard/ (3 layers):

• Builds clean
• Output is valid XML; each <g inkscape:groupmode="layer"> contains <path> elements directly — no nested `
• Layer z-order matches gerbv's normal rendering order (last_loaded → 0)
• Cairo does not emit <defs>/clip-path for Gerber content, so the string-extraction approach loses nothing in practice (though worth a comment if that ever changes)
• Pixel-for-pixel identical to the flat SVG export (ImageMagick compare -metric AE = 0)

Otherwise looks good — backward-compatible, feature works correctly.