summaryrefslogtreecommitdiff
path: root/src/selection_pipe.zig
diff options
context:
space:
mode:
authorGabriel Schneider <[email protected]>2026-10-23 17:10:58 -0300
committerGabriel Schneider <[email protected]>2026-10-23 17:10:58 -0300
commit0073d05329335dbb0eb85f5c3044e42805d144e5 (patch)
tree744edfc317719b2018b1c3a8657c223aec058d47 /src/selection_pipe.zig
parenta8b457a306a4442975210bcac3f04b56513c61c3 (diff)
downloadpardes-0073d05329335dbb0eb85f5c3044e42805d144e5.tar.gz
pardes-0073d05329335dbb0eb85f5c3044e42805d144e5.zip
Callstack lists every caller and pages a tree too big for one query; memory bounds and resident-size fixes
Callstack walks while the server answers, counts on the message row past a second, puts test callers last at each level, and ends a tree past one query's memory (memory.limits.arena.lsp_query, a 256 MiB region taken and given back per query) in a row whose Callstack shows the next page. Grep reads whole files. New hard limits: undo text per file, image decode pixels, the LSP snapshot of other open files, one paste. Safe builds no longer fill worst-case buffers with 0xAA (memory.unfilled), big regions are kept off transparent huge pages, and malloc_trim runs after big results. Also: Look's git show pages into a +Diff, and dropped marks in File. Co-Authored-By: Claude Opus 5.5 <[email protected]>
Diffstat (limited to 'src/selection_pipe.zig')
-rw-r--r--src/selection_pipe.zig11
1 files changed, 7 insertions, 4 deletions
diff --git a/src/selection_pipe.zig b/src/selection_pipe.zig
index c3e5e3a6..229b3d67 100644
--- a/src/selection_pipe.zig
+++ b/src/selection_pipe.zig
@@ -401,7 +401,7 @@ pub fn runIn(
},
};
child.stdin = null; // writer_context owns and closes this endpoint
- var writer: ?std.Thread = std.Thread.spawn(.{}, writeInput, .{writer_context}) catch {
+ var writer: ?std.Io.Future(void) = io.concurrent(writeInput, .{writer_context}) catch {
writer_context.file.close(io);
gpa.free(writer_context.input);
gpa.destroy(writer_context);
@@ -410,8 +410,9 @@ pub fn runIn(
};
// On every early return (a timeout, a ceiling) the answer goes at once:
// the group is killed and the shell left to be reaped elsewhere
- // (abandon), and the writer, which owns its input, let go.
- defer if (writer) |thread| thread.detach();
+ // (abandon), and the writer's write canceled (it closes and frees what
+ // it owns as it ends).
+ defer if (writer) |*w| w.cancel(io);
defer if (child.id != null) abandonNow(io, &child, pid);
defer if (token != 0) Running.remove(pid);
if (token != 0 and !Running.add(token, pid)) return fail.k(.signal);
@@ -443,7 +444,9 @@ pub fn runIn(
multi_reader.checkAnyError() catch return fail.k(.io);
if (token != 0) Running.remove(pid);
const term = child.wait(io) catch return fail.k(.io);
- writer.?.join();
+ // A cancel of this job reaches the write too: a grandchild that keeps
+ // stdin open and never reads holds no worker.
+ writer.?.await(io);
writer = null;
const stdout = multi_reader.toOwnedSlice(0) catch return fail.k(.io);