diff --git a/ext/iconv/tests/bug48147.phpt b/ext/iconv/tests/bug48147.phpt
index ce304eecfb3c7..bce73893350a4 100644
--- a/ext/iconv/tests/bug48147.phpt
+++ b/ext/iconv/tests/bug48147.phpt
@@ -2,6 +2,16 @@
 Bug #48147 (iconv with //IGNORE cuts the string)
 --EXTENSIONS--
 iconv
+--SKIPIF--
+<?php
+// POSIX 2024 standardizes "//IGNORE", but does NOT say that
+// it should ignore invalid input sequences, only untranslatable
+// ones. The GNU implementations have however historically skipped
+// invalid input sequences when it is used.
+if (ICONV_IMPL != "libiconv" && ICONV_IMPL != "glibc") {
+    die("skip iconv will not //IGNORE invalid input sequences");
+}
+?>
 --FILE--
 <?php
 $text = "aa\xC3\xC3\xC3\xB8aa";
diff --git a/ext/iconv/tests/bug52211.phpt b/ext/iconv/tests/bug52211.phpt
index f213c764de815..a68580447374d 100644
--- a/ext/iconv/tests/bug52211.phpt
+++ b/ext/iconv/tests/bug52211.phpt
@@ -1,5 +1,16 @@
 --TEST--
 Bug #52211 (iconv() returns part of string on error)
+--SKIPIF--
+<?php
+// The two GNU iconvs (glibc and libiconv) fail if the input sequence
+// cannot be converted to the target charset, but most others
+// (including musl, FreeBSD, NetBSD, and Solaris) do not. POSIX says
+// that "iconv() shall perform an implementation-defined conversion on
+// the character," so this is likely to remain a GNUism.
+if (ICONV_IMPL != "libiconv" && ICONV_IMPL != "glibc") {
+    die("skip iconv impl may not fail on unconvertable sequences");
+}
+?>
 --EXTENSIONS--
 iconv
 --FILE--
diff --git a/ext/iconv/tests/bug76249.phpt b/ext/iconv/tests/bug76249.phpt
index 37608ccc04470..39e549d7a7f75 100644
--- a/ext/iconv/tests/bug76249.phpt
+++ b/ext/iconv/tests/bug76249.phpt
@@ -7,13 +7,13 @@ iconv
 $fh = fopen('php://memory', 'rw');
 fwrite($fh, "abc");
 rewind($fh);
-if (false === @stream_filter_append($fh, 'convert.iconv.ucs-2/utf8//IGNORE', STREAM_FILTER_READ, [])) {
-    stream_filter_append($fh, 'convert.iconv.ucs-2/utf-8//IGNORE', STREAM_FILTER_READ, []);
+if (false === @stream_filter_append($fh, 'convert.iconv.ucs-2/utf8', STREAM_FILTER_READ, [])) {
+    stream_filter_append($fh, 'convert.iconv.ucs-2/utf-8', STREAM_FILTER_READ, []);
 }
 var_dump(stream_get_contents($fh));
 ?>
 DONE
 --EXPECTF--
-Warning: stream_get_contents(): iconv stream filter ("ucs-2"=>"utf%A8//IGNORE"): invalid multibyte sequence in %sbug76249.php on line %d
+Warning: stream_get_contents(): iconv stream filter ("ucs-2"=>"utf%A8"): invalid multibyte sequence in %sbug76249.php on line %d
 string(0) ""
 DONE
diff --git a/ext/iconv/tests/eucjp2iso2022jp.phpt b/ext/iconv/tests/eucjp2iso2022jp.phpt
index 6c1a8ec1214bc..c3260c370f08f 100644
--- a/ext/iconv/tests/eucjp2iso2022jp.phpt
+++ b/ext/iconv/tests/eucjp2iso2022jp.phpt
@@ -2,6 +2,16 @@
 EUC-JP to ISO-2022-JP
 --EXTENSIONS--
 iconv
+--SKIPIF--
+<?php
+// ISO-2022-JP is a stateful encoding, so the right answer is not
+// unique. In particular, musl (type "unknown") is known to have an
+// inefficient encoding for it that does not agree with the expected
+// output below.
+if (ICONV_IMPL == "unknown") {
+    die("skip byte-comparison of stateful encoding with unknown iconv");
+}
+?>
 --INI--
 error_reporting=2039
 --FILE--
diff --git a/ext/iconv/tests/iconv_mime_encode.phpt b/ext/iconv/tests/iconv_mime_encode.phpt
index e055f53b016ac..c6f41c56ad908 100644
--- a/ext/iconv/tests/iconv_mime_encode.phpt
+++ b/ext/iconv/tests/iconv_mime_encode.phpt
@@ -2,6 +2,16 @@
 iconv_mime_encode()
 --EXTENSIONS--
 iconv
+--SKIPIF--
+<?php
+// ISO-2022-JP is a stateful encoding, so the right answer is not
+// unique. In particular, musl (type "unknown") is known to have an
+// inefficient encoding for it that does not agree with the expected
+// output below.
+if (ICONV_IMPL == "unknown") {
+    die("skip byte-comparison of stateful encoding with unknown iconv");
+}
+?>
 --INI--
 iconv.internal_charset=iso-8859-1
 --FILE--