Repository navigation
Render: a curve with a point at infinity does not end the process - #224
Merged
Merged
Conversation
fonts/cjk supplies Noto Sans JP, SC and KR, variable fonts whose default is Thin, and the face was made of the default: text of a font a document names and does not embed was drawn in hairlines, whatever weight it named. The stroke of 「一」 at 100 points was three pixels. A program a FontProvider supplies that has a weight axis is used at the weight the document asks for: /FontWeight of the descriptor, else 700 for a bold font and 400 for any other, within what the axis has. The instance is baked once and the last eight are kept by program and weight, so the documents of a process share them where the provider hands out the same bytes. A font that cannot be instanced is used as it is, and an embedded variable font too: its writer meant the default. The stroke is eight pixels now, wider for a bold font and thinner for one of weight 100. Fixes #167.
A render with RenderOptions.Form draws a widget whose value is not the file's from a list of its own, after the page and all its annotations. An annotation that lies on the widget and follows it in /Annots (a line across a field, a stamp on it) was above the field as long as it showed the file's value and under it once the value changed. The page's list is now cut at such a widget: it is drawn up to the widget, then the widget, then the annotations that follow, up to the next widget something lies on. The list knows where its annotations begin and what each of them can touch, so that a pass after the first replays the clip to the page and the annotations of its range, in the part of the region they touch. A widget nothing lies on is an overlay over a page drawn in one piece, as before. The later passes draw onto the background, as the widgets do: a blend mode of such an annotation sees the background where the page drew nothing. Fixes #164
A page drawn in bands or tiles differed from the page drawn at once by one or two levels in a few antialiased pixels, and so did a page drawn by several workers from one drawn by one: of the first pages of 965 files of the pdf.js corpus at 72 dpi, 85. It was not the sampling of images, as the issue guessed, but the rasterizer: edges cut at the region and rounded there, runs and coverage blended with different rounding, analytic stroke rows counted from the region's first pixel (timzifer/stilus#36). stilus draws a pixel alike in every region now; this takes that version. TestRenderTiles draws a page that is all edges at once, in rows, columns and tiles, by two and five workers and in grey, and holds them against one another byte for byte. The corpus step "grey", which draws large pages in bands, allowed two levels and allows none now. Fixes #211
The release that has the change (timzifer/stilus#37), in place of its commit.
A coordinate too large for a single-precision number is infinite, and under the matrix of a page turned by a quarter infinity times zero is not a number. stilus flattened a curve with such a point without end (timzifer/stilus#38): the stack ran out, which ends the process and is no panic a render could recover. stilus v0.10.0, which this branch has, adds nothing for such a curve. TestCurveNotNumbers draws pages with such curves — filled, stroked, dashed, as a clip, in every rotation — and holds that the render returns and that what stands beside the curve is drawn.
timzifer
added
bug
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stacked on #223, which takes stilus v0.10.0.
The bug
A page could end the process that draws it. A coordinate of a curve that is too large for a single-precision number is infinite; under the matrix of a page turned by a quarter, infinity times zero is not a number; and stilus flattened a curve with such a point without end (timzifer/stilus#38). The stack ran out — a fatal error, not a panic, so
Renderdid not return a*PanicError: the process was gone.By
SECURITY.mdthis is of the kind that counts ("a panic that escapes", and worse). cera v0.10.0 and before are affected; the fix is in stilus v0.10.0.Here
TestCurveNotNumbers: pages with such curves —c,v,y; filled, stroked, dashed, as a clip; the large number as a control point, an end point, negative; in all four rotations — drawn by one and by three workers and read as text. The render returns, and the square beside the curve is drawn. With stilus v0.9.0 the test binary dies at the first turned page.Nothing else changes: the fix came with the stilus version of #223.