[wasm] Saturate Graphite positions

The standalone WebAssembly Graphite shaper mirrors the native backend
and had the same unchecked coordinate, cluster, and pen arithmetic.
Clamp backend values and saturate its signed position accumulation.

This is part of the complete position-overflow fix for issue #6151.

Testing: g++ -std=c++17 -Isrc -fsyntax-only src/wasm/graphite/shape.cc
Testing: meson test -C build --print-errorlogs

Assisted-by: Codex <codex@openai.com>
This commit is contained in:
Behdad Esfahbod committed 2026-08-12 18:27:23 -06:00
1 parent 3385a2deb0
commit f222478bcc
1 file changed
+42 -23
+42 -23
View File
@@ -7,6 +7,22 @@
#include <stdlib.h>
#include <string.h>
static int32_t
clamp_position (double value)
{
if (value <= INT32_MIN) return INT32_MIN;
if (value >= INT32_MAX) return INT32_MAX;
return (int32_t) value;
}
static int32_t
saturate_add (int32_t a, int32_t b)
{ return clamp_position ((int64_t) a + b); }
static int32_t
saturate_sub (int32_t a, int32_t b)
{ return clamp_position ((int64_t) a - b); }
void debugprint1 (char *s, int32_t);
void debugprint2 (char *s, int32_t, int32_t);
@@ -75,7 +91,7 @@ shape (void *shape_plan,
gr_segment *seg = nullptr;
const gr_slot *is;
unsigned int ci = 0, ic = 0;
unsigned int curradvx = 0, curradvy = 0;
int32_t curradvx = 0, curradvy = 0;
unsigned length = contents.length;
uint32_t *chars = (uint32_t *) malloc (length * sizeof (uint32_t));
@@ -103,7 +119,7 @@ shape (void *shape_plan,
unsigned int base_glyph;
unsigned int num_glyphs;
unsigned int cluster;
int advance;
int32_t advance;
};
length = glyph_count;
@@ -127,11 +143,11 @@ shape (void *shape_plan,
float xscale = (float) font_x_scale / upem;
float yscale = (float) font_y_scale / upem;
yscale *= yscale / xscale;
unsigned int curradv = 0;
int32_t curradv = 0;
if (DIRECTION_IS_BACKWARD (direction))
{
curradv = gr_slot_origin_X(gr_seg_first_slot(seg)) * xscale;
clusters[0].advance = gr_seg_advance_X(seg) * xscale - curradv;
curradv = clamp_position (gr_slot_origin_X(gr_seg_first_slot(seg)) * xscale);
clusters[0].advance = clamp_position ((double) (gr_seg_advance_X(seg) * xscale) - curradv);
}
else
clusters[0].advance = 0;
@@ -145,7 +161,7 @@ shape (void *shape_plan,
{
clusters[ci-1].num_chars += clusters[ci].num_chars;
clusters[ci-1].num_glyphs += clusters[ci].num_glyphs;
clusters[ci-1].advance += clusters[ci].advance;
clusters[ci-1].advance = saturate_add (clusters[ci-1].advance, clusters[ci].advance);
ci--;
}
@@ -159,14 +175,15 @@ shape (void *shape_plan,
c->num_glyphs = 0;
if (DIRECTION_IS_BACKWARD (direction))
{
c->advance = curradv - gr_slot_origin_X(is) * xscale;
curradv -= c->advance;
c->advance = clamp_position ((double) curradv - (double) (gr_slot_origin_X(is) * xscale));
curradv = saturate_sub (curradv, c->advance);
}
else
{
auto origin_X = gr_slot_origin_X (is) * xscale;
int32_t origin_X = clamp_position (gr_slot_origin_X (is) * xscale);
c->advance = 0;
clusters[ci].advance += origin_X - curradv;
clusters[ci].advance = saturate_add (clusters[ci].advance,
saturate_sub (origin_X, curradv));
curradv = origin_X;
}
ci++;
@@ -178,9 +195,10 @@ shape (void *shape_plan,
}
if (DIRECTION_IS_BACKWARD (direction))
clusters[ci].advance += curradv;
clusters[ci].advance = saturate_add (clusters[ci].advance, curradv);
else
clusters[ci].advance += gr_seg_advance_X(seg) * xscale - curradv;
clusters[ci].advance = clamp_position ((double) clusters[ci].advance +
(double) (gr_seg_advance_X(seg) * xscale) - curradv);
ci++;
for (unsigned int i = 0; i < ci; ++i)
@@ -204,37 +222,38 @@ shape (void *shape_plan,
curradvx = 0;
for (is = gr_seg_first_slot (seg); is; pPos++, ++info, is = gr_slot_next_in_segment (is))
{
pPos->x_offset = gr_slot_origin_X (is) * xscale - curradvx;
pPos->y_offset = gr_slot_origin_Y (is) * yscale - curradvy;
pPos->x_offset = clamp_position ((double) (gr_slot_origin_X (is) * xscale) - curradvx);
pPos->y_offset = clamp_position ((double) (gr_slot_origin_Y (is) * yscale) - curradvy);
if (info->cluster != currclus) {
pPos->x_advance = (int) info->var1;
curradvx += pPos->x_advance;
curradvx = saturate_add (curradvx, pPos->x_advance);
currclus = info->cluster;
} else
pPos->x_advance = 0.;
pPos->y_advance = gr_slot_advance_Y (is, grface, nullptr) * yscale;
curradvy += pPos->y_advance;
pPos->y_advance = clamp_position (gr_slot_advance_Y (is, grface, nullptr) * yscale);
curradvy = saturate_add (curradvy, pPos->y_advance);
}
buffer_set_contents (buffer, &contents);
}
else
{
curradvx = gr_seg_advance_X(seg) * xscale;
curradvx = clamp_position (gr_seg_advance_X(seg) * xscale);
for (is = gr_seg_first_slot (seg); is; pPos++, info++, is = gr_slot_next_in_segment (is))
{
if (info->cluster != currclus)
{
pPos->x_advance = (int) info->var1;
curradvx -= pPos->x_advance;
curradvx = saturate_sub (curradvx, pPos->x_advance);
currclus = info->cluster;
} else
pPos->x_advance = 0.;
pPos->y_advance = gr_slot_advance_Y (is, grface, nullptr) * yscale;
curradvy -= pPos->y_advance;
pPos->x_offset = gr_slot_origin_X (is) * xscale - (int) info->var1 - curradvx + pPos->x_advance;
pPos->y_offset = gr_slot_origin_Y (is) * yscale - curradvy;
pPos->y_advance = clamp_position (gr_slot_advance_Y (is, grface, nullptr) * yscale);
curradvy = saturate_sub (curradvy, pPos->y_advance);
pPos->x_offset = clamp_position ((double) (gr_slot_origin_X (is) * xscale) -
(int) info->var1 - curradvx + pPos->x_advance);
pPos->y_offset = clamp_position ((double) (gr_slot_origin_Y (is) * yscale) - curradvy);
}
buffer_set_contents (buffer, &contents);
buffer_reverse_clusters (buffer);