Fix incorrect lineJoin and pathfactor in get_drawings()

[sakamotch]

Fixes #4954

Changes

Bug 1: lineJoin scaled by pathfactor

stroke->linejoin is an enum (0=Miter, 1=Round, 2=Bevel) but was multiplied by pathfactor. Changed to emit the raw integer value, consistent with how lineCap is already handled.

// Before
Py_BuildValue("f", dev->pathfactor * stroke->linejoin)

// After
Py_BuildValue("i", stroke->linejoin)

Bug 2: pathfactor falls back to 1 for non-uniform scaling

The original calculation only covered uniform scaling (|a|==|d|) and 90° rotation (|b|==|c|). Changed to use sqrt(a² + b²) which handles arbitrary transforms.

// Before
dev->pathfactor = 1;
if (ctm.a != 0 && fz_abs(ctm.a) == fz_abs(ctm.d))
    dev->pathfactor = fz_abs(ctm.a);
else if (ctm.b != 0 && fz_abs(ctm.b) == fz_abs(ctm.c))
    dev->pathfactor = fz_abs(ctm.b);

// After
float scale = sqrtf(ctm.a * ctm.a + ctm.b * ctm.b);
if (scale < 1e-9f)
    scale = sqrtf(ctm.c * ctm.c + ctm.d * ctm.d);
if (scale < 1e-9f)
    scale = 1.0f;
dev->pathfactor = scale;

Note: For non-uniform scaling, stroke width is direction-dependent in general. This fix approximates it using the length of the transformed unit vector.

Files changed

• src/extra.i — rebased build
• src_classic/helper-devices.i — classic build
• src/__init__.py — Python fallback

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[sakamotch]

I have read the CLA Document and I hereby sign the CLA

[julian-smith-artifex-com]

Many thanks for updating things.

Could your squash your 4 commits into 1 commit and rebase on top of current main?

Then i think i'd be ok to merge this PR, after checking with @JorjMcKie.

[sakamotch]

Thanks for the review.
Squashed into one commit and rebased onto current main.

[sakamotch]

I checked the CI failure more carefully. The determinant-based pathfactor I introduced here is the problem.

I'll change it to:

float sx = sqrtf(ctm.a * ctm.a + ctm.b * ctm.b);
float sy = sqrtf(ctm.c * ctm.c + ctm.d * ctm.d);
float dot = ctm.a * ctm.c + ctm.b * ctm.d;

if (sx == sy && dot == 0.0f)
    dev->pathfactor = sx;
else
    dev->pathfactor = 1.0f;

This extends the old logic from 0/90-degree cases to arbitrary rotation angles, while keeping the same strict fallback behavior.

I’m keeping == intentionally, since the previous implementation already used exact float comparisons.

I tested this locally and tests/test_annots.py::test_redact4 now passes, and tests/test_drawings.py also passes with this change.

[sakamotch]

@julian-smith-artifex-com If this revised approach looks good, I can update the PR with a new squashed commit.

[julian-smith-artifex-com]

test_redact4 is currently failing due to a regression in mupdf in the latest master branch. So i don't think you should try to get the test to pass here.

Tests are passing with this PR and git branch 1.27.x, so i think the PR is good.

[julian-smith-artifex-com]

[rerunning tests with mupdf master as failure was from mupdf, not thie PR.]

[julian-smith-artifex-com]

I hope you don't mind, but i've taken the liberty of rebasing the PR tree on top of latest main.

Hopefully it will still pass tests, and we can merge the PR.

[sakamotch]

Thanks again for taking care of the earlier rebase.

It looks like main moved forward again, so I rebased the PR branch onto the latest main. Hope that's okay, and hopefully it should be mergeable now.

[sakamotch]

@julian-smith-artifex-com Checks passed, looks mergeable now!