motiejus/zig
1
Fork 0
Code Releases Activity

sema: fix shard simulation, bump num_passing 61→67

Browse Source 
Fix the IP shard simulation to work for all main analysis, not just
during ensureFullMemoizedStateC. This fixes tests 61-66 (inline fn
tests and non-inline tests that reference types from preamble builtins).

Key changes:
- Move skip_dedup activation to the end of the preamble (after
  analyzeMemoizedStateC + start.zig comptime). Previously it was only
  set during ensureFullMemoizedStateC, missing tests that don't
  trigger full builtin resolution.
- Expand cc_keep range by 1 entry to include the CC ptr_nav that
  AddressSpace resolution creates as a side effect (fixing the
  1-entry excess for inline function tests).
- Remove skip_dedup clearing from ensureFullMemoizedStateC (it stays
  active for the entire main analysis).

Next blocker: test 67 (shl_sat.zig) — the shift amount type (u6 for
u32 operand) is not created. C stores the shift amount as u32(1)
instead of u6(1). Need to port shift amount type coercion.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Motiejus Jakštys
2026-03-02 17:54:04 +00:00
parent 6390ee4b8a
commit 69a0486232
2 changed files with 25 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
stage0/corpus.zig
View File

@@ -3,7 +3,7 @@
/// `num_passing` controls how many files are tested and pre-generated.
/// Both build.zig and stages_test.zig import this file.
/// To enable more tests: just increment `num_passing`.
pub const num_passing: usize = 61;
pub const num_passing: usize = 67;
pub const files = [_][]const u8{
"stage0/sema_tests/empty.zig",

41
stage0/sema.c
View File

@@ -10551,8 +10551,13 @@ static void analyzeMemoizedStateC(Sema* sema) {
// Track CC (builtin 2) range: these entries are in Zig's local
// shard and should NOT be skipped during main analysis dedup.
// Start 1 entry earlier to include the CC ptr_nav that
// AddressSpace resolution creates as a side effect (the ptr_nav
// for the CallingConvention declaration is created when the
// std.builtin namespace is scanned during AddressSpace resolution).
if (i == 2)
sema->zcu->preamble_cc_start = sema->ip->items_len;
sema->zcu->preamble_cc_start
= sema->ip->items_len > 0 ? sema->ip->items_len - 1 : 0;
InternPoolIndex val = ensureNavValUpToDate(sema, nav);
if (val == IP_INDEX_NONE)
@@ -10597,18 +10602,12 @@ static void ensureFullMemoizedStateC(Sema* sema) {
= sema->zcu->file_namespaces[sema->zcu->builtin_file_idx];
// Shard simulation: clear builtins 0-4 state so they get re-resolved
// as fresh entries during main analysis. Ported from upstream: Zig's
// sharded IP keeps preamble entries in a separate shard, so main
// analysis creates fresh entries for builtins 0-4 types. We simulate
// this by enabling skip_dedup on the preamble memoized range and
// clearing the cached nav resolutions.
// as fresh entries during main analysis. The skip_dedup range is
// already set at preamble end (see preamble setup). Here we just
// clear the cached nav resolutions so ensureNavValUpToDate re-resolves
// builtins 0-4 with fresh IP entries.
if (sema->zcu->preamble_memoized_end
> sema->zcu->preamble_memoized_start) {
sema->ip->skip_dedup_start = sema->zcu->preamble_memoized_start;
sema->ip->skip_dedup_end = sema->zcu->preamble_memoized_end;
sema->ip->cc_keep_start = sema->zcu->preamble_cc_start;
sema->ip->cc_keep_end = sema->zcu->preamble_cc_end;
// Clear builtins 0-4 values so they're re-resolved.
for (int i = 0; i < 5 && i < NUM_BUILTIN_DECL_MAIN; i++)
sema->zcu->builtin_decl_values[i] = IP_INDEX_NONE;
@@ -10653,12 +10652,6 @@ static void ensureFullMemoizedStateC(Sema* sema) {
sema->zcu->builtin_decl_values[i] = val;
}
// Disable shard simulation after Phase 1.
sema->ip->skip_dedup_start = 0;
sema->ip->skip_dedup_end = 0;
sema->ip->cc_keep_start = 0;
sema->ip->cc_keep_end = 0;
// Phase 2: Call resolveTypeFullyC on ALL type builtins.
// Ported from analyzeMemoizedState line 37560:
// try uncoerced_val.toType().resolveFully(pt);
@@ -12788,6 +12781,20 @@ SemaFuncAirList semaAnalyze(Sema* sema) {
sema->zcu->in_main_analysis = false;
}
// Enable shard simulation for the entire main analysis.
// Ported from upstream: Zig's sharded IP keeps preamble
// memoized state entries in a separate shard. Any main
// analysis interning that matches a preamble entry will
// create a fresh entry instead of deduplicating.
if (sema->zcu->preamble_memoized_end
> sema->zcu->preamble_memoized_start) {
sema->ip->skip_dedup_start
= sema->zcu->preamble_memoized_start;
sema->ip->skip_dedup_end = sema->zcu->preamble_memoized_end;
sema->ip->cc_keep_start = sema->zcu->preamble_cc_start;
sema->ip->cc_keep_end = sema->zcu->preamble_cc_end;
}
} else {
// No module root — root module is file_idx=0.
if (sema->zcu->num_files == 0) {