7a356d85dba1647c6d918b315e075da7e31ab7cb braney Sat Sep 26 12:10:36 2026 -0700 htmlSanitize: keep simple svg drawings in hub description pages, refs #38428 Description pages that draw small svg shapes, such as a colour legend, came out empty. svg, g, circle, ellipse, rect, line, polyline, polygon and path now come through with their size, position, fill and stroke attributes. Other svg elements are left out. Over the 5390-page hub corpus only the 32 pages that contain an svg change, and no page's text changes. diff --git src/lib/tests/htmlSanitizeTest.c src/lib/tests/htmlSanitizeTest.c index cd081f855cd..6d4595e3670 100644 --- src/lib/tests/htmlSanitizeTest.c +++ src/lib/tests/htmlSanitizeTest.c @@ -25,56 +25,77 @@ /* ordinary links are left alone, and a new window does not get a handle on ours */ "u a", /* an image keeps its source, not its onerror */ "\"a\"", /* a frame survives only when it plays a video from a host we know */ "" "", /* the style attribute is filtered a property at a time */ "

p

", /* unknown elements lose their tag and keep their text */ "wordmore", /* tags left open are closed for us, and a slash does not close one of these */ "
bold

para", "

the rest of the page is not inside that div", /* a page built of tags that are never closed is not a way to make us work all day */ -"text", +"text", /* a stray quote inside an unquoted value does not swallow the page */ "link text and more text", /* a form and everything in it goes */ "

after

", /* comments and doctypes go */ "

shown

", /* an id, and a name on an anchor, are renamed, and a link to one of them is renamed too */ "

M

t" "same page other page" "to the top
no name at all
", /* a table keeps its shape */ "
cell
", /* an entity in a style value cannot spell a property value we would not print */ "

a

", /* an ampersand the author meant still reads as one */ "

a

", /* the same attribute twice is written once, the way a browser reads it */ "\"a\"" "x", /* a quote that is never closed takes the rest of the page, and we say so */ "

real

never printed

", /* a less than sign that starts no tag is text, and cannot eat a tag of ours */ "
a < bbold
", /* a name we already renamed is not renamed again, on either side of the link */ "

M

jump", +/* a colored circle in a legend survives, the way UniBind draws one */ +"\n" + "", +/* shapes written with a closing slash are siblings, not each inside the one before */ +"" + "", +/* nothing in a drawing may run, link, load or animate */ +"" + "" + "" + "" + "

text

", +/* a fill or a stroke cannot fetch anything, however the url is spelled */ +"" + "", +/* a shape that only clips another is not drawn on its own */ +"" + "", +/* a shape outside a drawing loses its tag */ +"text after", }; int main(int argc, char *argv[]) { int i; for (i = 0; i < ArraySize(cases); ++i) { char *clean = htmlSanitize(cases[i]); printf("in : %s\nout: %s\n", cases[i], clean); struct slName *removed = NULL, *el; freeMem(clean); clean = htmlSanitizeReport(cases[i], &removed); for (el = removed; el != NULL; el = el->next) printf(" (%s)\n", el->name); printf("\n");