Skip to content

fix(scanner): avoid GCC -O2 strict-aliasing miscompile in scan_start_tag #2

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
swaits wants to merge 1 commit into
base: main
Choose a base branch
from
Open

fix(scanner): avoid GCC -O2 strict-aliasing miscompile in scan_start_tag #2

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
24 changes: 23 additions & 1 deletion crates/tree-sitter-htmlx/src/scanner.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -345,7 +345,29 @@ static bool scan_start_tag(State *state, TSLexer *lexer, const bool *valid) {
lexer->mark_end(lexer);
}
Tag tag = htmlx_tag_for_svelte_name(name, saw_ascii_upper);
array_push(&state->html->tags, tag);

// Append onto the HTML scanner's tag stack.
//
// This is done explicitly rather than with
// `array_push(&state->html->tags, tag)`. array_push() reallocates the
// backing buffer through array.h's generic `Array *` type, then writes
// the new element back through the concrete `Array(Tag) *` type. When
// the array is reached through the `state->html` pointer, GCC's
// strict-aliasing analysis at -O2 treats the realloc store and the
// element store as non-aliasing and elides the reload of `contents` —
// leaving a stale NULL pointer and segfaulting on the first tag of
// every document. Keeping every access on the concrete `Array(Tag)`
// type avoids the miscompile.
Scanner *html = state->html;
if (html->tags.size + 1 > html->tags.capacity) {
uint32_t new_capacity =
html->tags.capacity ? html->tags.capacity * 2 : 8;
html->tags.contents = (Tag *)ts_realloc(
html->tags.contents, (size_t)new_capacity * sizeof(Tag));
html->tags.capacity = new_capacity;
}
html->tags.contents[html->tags.size++] = tag;

state->open_tag_is_namespaced = false;

switch (tag.type) {
Expand Down
24 changes: 23 additions & 1 deletion crates/tree-sitter-svelte/src/htmlx/scanner.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -349,7 +349,29 @@ static bool scan_start_tag(State *state, TSLexer *lexer, const bool *valid) {
lexer->mark_end(lexer);
}
Tag tag = htmlx_tag_for_svelte_name(name, saw_ascii_upper);
array_push(&state->html->tags, tag);

// Append onto the HTML scanner's tag stack.
//
// This is done explicitly rather than with
// `array_push(&state->html->tags, tag)`. array_push() reallocates the
// backing buffer through array.h's generic `Array *` type, then writes
// the new element back through the concrete `Array(Tag) *` type. When
// the array is reached through the `state->html` pointer, GCC's
// strict-aliasing analysis at -O2 treats the realloc store and the
// element store as non-aliasing and elides the reload of `contents` —
// leaving a stale NULL pointer and segfaulting on the first tag of
// every document. Keeping every access on the concrete `Array(Tag)`
// type avoids the miscompile.
Scanner *html = state->html;
if (html->tags.size + 1 > html->tags.capacity) {
uint32_t new_capacity =
html->tags.capacity ? html->tags.capacity * 2 : 8;
html->tags.contents = (Tag *)ts_realloc(
html->tags.contents, (size_t)new_capacity * sizeof(Tag));
html->tags.capacity = new_capacity;
}
html->tags.contents[html->tags.size++] = tag;

state->open_tag_is_namespaced = false;

switch (tag.type) {
Expand Down