Skip to content

sapi/litespeed: Stop ub_write reporting a failed write as complete - #24124

Open
iliaal wants to merge 1 commit into
php:PHP-8.4from
iliaal:fix/aph-litespeed-write-accounting-opzz-84-work
Open

iliaal wants to merge 1 commit into
php:PHP-8.4from
iliaal:fix/aph-litespeed-write-accounting-opzz-84-work

Conversation

@iliaal

@iliaal iliaal commented Oct 4, 2026

Copy link
Copy Markdown
Member

Picks up #23426, which fixed the short-write return value and was closed for inactivity. LSAPI_Write() only ever returns the full length or -1, and the -1 was the bigger problem: compared as an int against a size_t length it never took the error branch, so a failed write counted as complete. LiteSpeed is not built in CI; the file compiles cleanly with -Wsign-compare.

Fixes #23424

LSAPI_Write() returns either the full length or -1, but
sapi_lsapi_ub_write() kept the result in an int and compared it with the
size_t length, so -1 never took the error branch: a failed write was
reported as complete and php_handle_aborted_connection() was not called.
The short-write branch also returned str_length - ret instead of ret.
The stdout fallback loop now counts in size_t and ssize_t as well.

Fixes phpGH-23424
@iluuu1994

Copy link
Copy Markdown
Member

@iliaal Did you test locally with litespeed?

@iliaal

iliaal commented Oct 5, 2026

Copy link
Copy Markdown
Member Author

Yes, with OpenLiteSpeed 1.9.2 running lsphp in LSAPI mode, streaming 64 KB chunks and dropping the client mid-response. Without the patch connection_aborted() stays 0 after the failed write and the script runs to the end; with it the abort is flagged and the script stops. Full responses are identical on both builds.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants