fork of https://github.com/sourcegraph/zoekt
fix: binary file check misses binary files where first byte is null (#1054)
While searching for a short string of text, it returned some TTF files.
I downloaded one from the source repo and checked it locally with a hex
viewer to confirm. A bit of poking around this logic bug.
I found a second more recent instance of the same bug that seems to be
a copy/paste error.
author
Jay Graves
committer
GitHub
date
(May 4, 2026, 8:01 AM +0200)
commit
5baf7e20
5baf7e20587e1b6ad0195742e439494b91aa1cb7
parent
f59b17b6
f59b17b6437b8546a5dcac42a92e31b645be9d8c
+4
-4
+2
-2
index/index_test.go
+2
-2
index/index_test.go
···
3409
3409
t.Errorf("Check(%q): %v", text, skip)
3410
3410
}
3411
3411
}
3412
-
for _, text := range []string{"zero\x00byte", "xx", "0123456789abcdefghi"} {
3412
+
for _, text := range []string{"zero\x00byte", "\x00starts with null byte", "xx", "0123456789abcdefghi"} {
3413
3413
if skip := docChecker.Check([]byte(text), 15, false); skip == SkipReasonNone {
3414
3414
t.Errorf("Check(%q) succeeded", text)
3415
3415
}
···
3421
3421
t.Errorf("Check(%q): %v", text, skip)
3422
3422
}
3423
3423
}
3424
-
for _, text := range []string{"zero\x00byte", "xx"} {
3424
+
for _, text := range []string{"zero\x00byte", "\x00starts with null byte", "xx"} {
3425
3425
if skip := docChecker.Check([]byte(text), 15, true); skip == SkipReasonNone {
3426
3426
t.Errorf("Check(%q) succeeded", text)
3427
3427
}
+2
-2
index/shard_builder.go
+2
-2
index/shard_builder.go
···
546
546
// Skip binary check if already computed (e.g., by Builder.Add
547
547
// which calls DocChecker.Check before docs reach buildShard).
548
548
if doc.Category == FileCategoryMissing {
549
-
if index := bytes.IndexByte(doc.Content, 0); index > 0 {
549
+
if index := bytes.IndexByte(doc.Content, 0); index >= 0 {
550
550
doc.SkipReason = SkipReasonBinary
551
551
}
552
552
}
···
686
686
return SkipReasonTooSmall
687
687
}
688
688
689
-
if index := bytes.IndexByte(content, 0); index > 0 {
689
+
if index := bytes.IndexByte(content, 0); index >= 0 {
690
690
return SkipReasonBinary
691
691
}
692
692