eucjp_2004_encoding.phpt fails on macOS in release builds

Description

https://github.com/php/php-src/actions/runs/20119044115/job/57735104285

Given this succeeds in debug builds, this might be:

Undefined behavior in php-src
A miscompilation in (Apple) Clang

UBSan did not report any issues, though that doesn't mean much.

Reduced excerpt

006+ mb_check_encoding failed on good EUC-JP-2004 string: (8ff4df8ffed5afe0c7e6cfc5edd8e8cdcbbfdef0d6ee)
007+ mb_convert_encoding not working on EUC-JP-2004 input: (8ff4df8ffed5afe0c7e6cfc5edd8e8cdcbbfdef0d6ee)
008+ Expected UTF-32BE: (00007e9800009e850002146e0000966a00008cc400008f4200008422000067d000006ea500005e57)
009+ Actually got:      (00007e9800009e85000000250000966a00008cc400008f4200008422000067d000006ea500005e57)
010+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (8ff4df8ffed5afe0c7e6cfc5edd8e8cdcbbfdef0d6ee) when converting to UTF-32BE
011+ mb_check_encoding failed on good EUC-JP-2004 string: (d0e8adb4e0fbabc0bcddf6a9afa2)
012+ mb_convert_encoding not working on EUC-JP-2004 input: (d0e8adb4e0fbabc0bcddf6a9afa2)
013+ Expected UTF-32BE: (00004ff600002473000074ca00000291000067530002373f000054c6)
014+ Actually got:      (00004ff600002473000074ca000002910000675300000025000054c6)
015+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (d0e8adb4e0fbabc0bcddf6a9afa2) when converting to UTF-32BE
016+ mb_check_encoding failed on good EUC-JP-2004 string: (8fa8d9cfeed6d0d3e9f7e68ff8aaefeed6e4)
017+ mb_convert_encoding not working on EUC-JP-2004 input: (8fa8d9cfeed6d0d3e9f7e68ff8aaefeed6e4)
018+ Expected UTF-32BE: (000037f400021e3400005d840000569400003e3f00027614000095d400005e1a)
019+ Actually got:      (000037f40000002500005d840000569400003e3f00027614000095d400005e1a)
020+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (8fa8d9cfeed6d0d3e9f7e68ff8aaefeed6e4) when converting to UTF-32BE
021+ mb_check_encoding failed on good EUC-JP-2004 string: (fdb98ff2f08fafec8fadb2e7a4e7c7faddf2ad)
022+ mb_convert_encoding not working on EUC-JP-2004 input: (fdb98ff2f08fafec8fadb2e7a4e7c7faddf2ad)
023+ Expected UTF-32BE: (00009348000079c700006af100022b4600008153000081df00026aff00009b2e)
024+ Actually got:      (00009348000079c700006af100022b4600008153000081df0000002500009b2e)
025+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (fdb98ff2f08fafec8fadb2e7a4e7c7faddf2ad) when converting to UTF-32BE
026+ mb_check_encoding failed on good EUC-JP-2004 string: (bde6c7c98fa5ace3cd8ff0eeb9d2fbb3a6d4)
027+ mb_convert_encoding not working on EUC-JP-2004 input: (bde6c7c98fa5ace3cd8ff0eeb9d2fbb3a6d4)
028+ Expected UTF-32BE: (0000918700006d3e0000588f00007a37000074240000822a00026e40000003c5)
029+ Actually got:      (0000918700006d3e0000588f00007a37000074240000822a00000025000003c5)
030+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (bde6c7c98fa5ace3cd8ff0eeb9d2fbb3a6d4) when converting to UTF-32BE
031+ mb_check_encoding failed on good EUC-JP-2004 string: (c6edf9a98ff8bcb9f0a1aed3f98ff9f3f9b7a4a8fcb8)
032+ mb_convert_encoding not working on EUC-JP-2004 input: (c6edf9a98ff8bcb9f0a1aed3f98ff9f3f9b7a4a8fcb8)
033+ Expected UTF-32BE: (000075770002550e0002770e0000544a0000ff40000056ee00008fe00000fa4e000030480000fa64)
034+ Actually got:      (00007577000000250002770e0000544a0000ff40000056ee00008fe00000fa4e000030480000fa64)
035+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (c6edf9a98ff8bcb9f0a1aed3f98ff9f3f9b7a4a8fcb8) when converting to UTF-32BE
036+ mb_check_encoding failed on good EUC-JP-2004 string: (f8c8f5f08feeefddaed3a8f6e08fa1c0dfd8c6fe)
037+ mb_convert_encoding not working on EUC-JP-2004 input: (f8c8f5f08feeefddaed3a8f6e08fa1c0dfd8c6fe)
038+ Expected UTF-32BE: (000075170000690a00006e2200006aea000054e600023cfe00004f16000070f100005165)
039+ Actually got:      (000075170000690a00006e2200006aea000054e60000002500004f16000070f100005165)
040+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (f8c8f5f08feeefddaed3a8f6e08fa1c0dfd8c6fe) when converting to UTF-32BE
041+ mb_check_encoding failed on good EUC-JP-2004 string: (8ff5e2a5fa8ffbd7e2e1e3f58ffaf2afc2b1f3abb3abeb)
042+ mb_convert_encoding not working on EUC-JP-2004 input: (8ff5e2a5fa8ffbd7e2e1e3f58ffaf2afc2b1f3abb3abeb)
043+ Expected UTF-32BE: (00008222000030b10000309a00028d10000077da00007b0a000289bb0002123d00009060000002500000031f)
044+ Actually got:      (00008222000030b10000309a00028d10000077da00007b0a000289bb0000002500009060000002500000031f)
045+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (8ff5e2a5fa8ffbd7e2e1e3f58ffaf2afc2b1f3abb3abeb) when converting to UTF-32BE
046+ mb_check_encoding failed on good EUC-JP-2004 string: (8ffca3dade8fafbc8ffcb1fca2dcaee1f1fbc9bac8)
047+ mb_convert_encoding not working on EUC-JP-2004 input: (8ffca3dade8fafbc8ffcb1fca2dcaee1f1fbc9bac8)
048+ Expected UTF-32BE: (000292b100006772000069c0000097f100008a370000692a0000763b000270f4000054c9)
049+ Actually got:      (000292b100006772000069c0000097f100008a370000692a0000763b00000025000054c9)
050+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (8ffca3dade8fafbc8ffcb1fca2dcaee1f1fbc9bac8) when converting to UTF-32BE
051+ mb_check_encoding failed on good EUC-JP-2004 string: (a6e5aea2f8e3baa1f6c5)
052+ mb_convert_encoding not working on EUC-JP-2004 input: (a6e5aea2f8e3baa1f6c5)
053+ Expected UTF-32BE: (000026170002000b0000769e00006b6400006b77)
054+ Actually got:      (00002617000000250000769e00006b6400006b77)
055+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (a6e5aea2f8e3baa1f6c5) when converting to UTF-32BE
056+ mb_check_encoding failed on good EUC-JP-2004 string: (bceecfe3f6cb8facd0f7c3)
057+ mb_convert_encoding not working on EUC-JP-2004 input: (bceecfe3f6cb8facd0f7c3)
058+ Expected UTF-32BE: (000073e0000216b400006bd6000226f300007064)
059+ Actually got:      (000073e00000002500006bd6000226f300007064)
060+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (bceecfe3f6cb8facd0f7c3) when converting to UTF-32BE
061+ mb_check_encoding failed on good EUC-JP-2004 string: (c3f28ff5a6f8da8ff4a1f8fea6e8a7aa8ff3d0)
062+ mb_convert_encoding not working on EUC-JP-2004 input: (c3f28ff5a6f8da8ff4a1f8fea6e8a7aa8ff3d0)
063+ Expected UTF-32BE: (000092f3000080110000764b00007ccd0002548e000026000000041800007b69)
064+ Actually got:      (000092f3000080110000764b00007ccd00000025000026000000041800007b69)
065+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (c3f28ff5a6f8da8ff4a1f8fea6e8a7aa8ff3d0) when converting to UTF-32BE
066+ mb_check_encoding failed on good EUC-JP-2004 string: (8ffaad2b8fa4cadbf4b9e9afcc8fa8c9cdc1a2e3b8a2)
067+ mb_convert_encoding not working on EUC-JP-2004 input: (8ffaad2b8fa4cadbf4b9e9afcc8fa8c9cdc1a2e3b8a2)
068+ Expected UTF-32BE: (000090dc0000002b000036050000688d000062f70002131b00005cc900008f3f0000226a00006a29)
069+ Actually got:      (000090dc0000002b000036050000688d000062f70000002500005cc900008f3f0000226a00006a29)
070+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (8ffaad2b8fa4cadbf4b9e9afcc8fa8c9cdc1a2e3b8a2) when converting to UTF-32BE
071+ mb_check_encoding failed on good EUC-JP-2004 string: (b9e5e5b2cdc3e9a1e9febaa7f9ee8fefaf)
072+ mb_convert_encoding not working on EUC-JP-2004 input: (b9e5e5b2cdc3e9a1e9febaa7f9ee8fefaf)
073+ Expected UTF-32BE: (000052ab00007d4f000050ad000085410000875900005a5a00025da100006f25)
074+ Actually got:      (000052ab00007d4f000050ad000085410000875900005a5a0000002500006f25)
075+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (b9e5e5b2cdc3e9a1e9febaa7f9ee8fefaf) when converting to UTF-32BE
076+ mb_check_encoding failed on good EUC-JP-2004 string: (8faffee6c9f1af8ff1a5ddd6d5d3f5f2)
077+ mb_convert_encoding not working on EUC-JP-2004 input: (8faffee6c9f1af8ff1a5ddd6d5d3f5f2)
078+ Expected UTF-32BE: (00006b8100007fe900009912000074ab00006bb700005b51000235c4)
079+ Actually got:      (00006b8100007fe900009912000074ab00006bb700005b5100000025)
080+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (8faffee6c9f1af8ff1a5ddd6d5d3f5f2) when converting to UTF-32BE
081+ mb_check_encoding failed on good EUC-JP-2004 string: (bfeff7ecc8acbab9edc58ff7add5eaa4a2)
082+ mb_convert_encoding not working on EUC-JP-2004 input: (bfeff7ecc8acbab9edc58ff7add5eaa4a2)
083+ Expected UTF-32BE: (0000968f000247f10000516b00005dee00008efc0000860000005be200003042)
084+ Actually got:      (0000968f000000250000516b00005dee00008efc0000860000005be200003042)
085+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (bfeff7ecc8acbab9edc58ff7add5eaa4a2) when converting to UTF-32BE
086+ mb_check_encoding failed on good EUC-JP-2004 string: (8ff5fa8ff9e9e5cef7c2cfd4feadedf3e0f0)
087+ mb_convert_encoding not working on EUC-JP-2004 input: (8ff5fa8ff9e9e5cef7c2cfd4feadedf3e0f0)
088+ Expected UTF-32BE: (000082830000fa6600007de10000704e00020b9f00009a030000901100007441)
089+ Actually got:      (000082830000fa6600007de10000704e0000002500009a030000901100007441)
090+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (8ff5fa8ff9e9e5cef7c2cfd4feadedf3e0f0) when converting to UTF-32BE
091+ mb_check_encoding failed on good EUC-JP-2004 string: (f9d4e7f6debeaafcc9dbbdb5baf5)
092+ mb_convert_encoding not working on EUC-JP-2004 input: (f9d4e7f6debeaafcc9dbbdb5baf5)
093+ Expected UTF-32BE: (000259c40000839300006db50000028100005e030000903100007a84)
094+ Actually got:      (000000250000839300006db50000028100005e030000903100007a84)
095+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (f9d4e7f6debeaafcc9dbbdb5baf5) when converting to UTF-32BE
096+ mb_check_encoding failed on good EUC-JP-2004 string: (f7dceef7dfeabdb1fbecceb1abd9)
097+ mb_convert_encoding not working on EUC-JP-2004 input: (f7dceef7dfeabdb1fbecceb1abd9)
098+ Expected UTF-32BE: (00007187000092cf0000714c0000897200027684000075590000030b)
099+ Actually got:      (00007187000092cf0000714c0000897200000025000075590000030b)
100+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (f7dceef7dfeabdb1fbecceb1abd9) when converting to UTF-32BE
101+ mb_check_encoding failed on good EUC-JP-2004 string: (d9ddcdf3fcd1c2f5fbd1e6ac)
102+ mb_convert_encoding not working on EUC-JP-2004 input: (d9ddcdf3fcd1c2f5fbd1e6ac)
103+ Expected UTF-32BE: (0000639f00006b04000283cd00006fef0000459d00007f68)
104+ Actually got:      (0000639f00006b040000002500006fef0000459d00007f68)
105+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (d9ddcdf3fcd1c2f5fbd1e6ac) when converting to UTF-32BE
106+ mb_check_encoding failed on good EUC-JP-2004 string: (eca6b8c78ff2b8c2b3dedad7bcbdd7b5c7ebd5fee6)
107+ mb_convert_encoding not working on EUC-JP-2004 input: (eca6b8c78ff2b8c2b3dedad7bcbdd7b5c7ebd5fee6)
108+ Expected UTF-32BE: (00008b80000056fa000077ec00007d9a00006e4d00005f4800007ae3000097a000008aa50002a190)
109+ Actually got:      (00008b80000056fa000077ec00007d9a00006e4d00005f4800007ae3000097a000008aa500000025)
110+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (eca6b8c78ff2b8c2b3dedad7bcbdd7b5c7ebd5fee6) when converting to UTF-32BE
111+ mb_check_encoding failed on good EUC-JP-2004 string: (dcd7f3ceada98ff9def6b2b1bffaeee7da)
112+ mb_convert_encoding not working on EUC-JP-2004 input: (dcd7f3ceada98ff9def6b2b1bffaeee7da)
113+ Expected UTF-32BE: (0000699500009ea500002468000282f3000237630000904b0000832200008259)
114+ Actually got:      (0000699500009ea500002468000282f3000000250000904b0000832200008259)
115+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (dcd7f3ceada98ff9def6b2b1bffaeee7da) when converting to UTF-32BE
116+ mb_check_encoding failed on good EUC-JP-2004 string: (e3a3b8d9a6bfcde1d7bbf5baccb8)
117+ mb_convert_encoding not working on EUC-JP-2004 input: (e3a3b8d9a6bfcde1d7bbf5baccb8)
118+ Expected UTF-32BE: (000078fd00008de80000266700006d7400005f41000231c400009727)
119+ Actually got:      (000078fd00008de80000266700006d7400005f410000002500009727)
120+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (e3a3b8d9a6bfcde1d7bbf5baccb8) when converting to UTF-32BE
121+ mb_check_encoding failed on good EUC-JP-2004 string: (bda5affbc7b8ebfeaecbeec58ffabb)
122+ mb_convert_encoding not working on EUC-JP-2004 input: (bda5affbc7b8ebfeaecbeec58ffabb)
123+ Expected UTF-32BE: (00006101000218bd000057dc00008b6b000051030000915600009137)
124+ Actually got:      (0000610100000025000057dc00008b6b000051030000915600009137)
125+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (bda5affbc7b8ebfeaecbeec58ffabb) when converting to UTF-32BE
126+ mb_check_encoding failed on good EUC-JP-2004 string: (c2ed8ffee8f9c7afddfed8d2af)
127+ mb_convert_encoding not working on EUC-JP-2004 input: (c2ed8ffee8f9c7afddfed8d2af)
128+ Expected UTF-32BE: (0000702700009f41000257710000589e00009d1d000052f5)
129+ Actually got:      (0000702700009f41000000250000589e00009d1d000052f5)
130+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (c2ed8ffee8f9c7afddfed8d2af) when converting to UTF-32BE
131+ mb_check_encoding failed on good EUC-JP-2004 string: (f4c6bdd2bbf0e0e3bdb4fcc9ddb48ff0a1f9f0)
132+ mb_convert_encoding not working on EUC-JP-2004 input: (f4c6bdd2bbf0e0e3bdb4fcc9ddb48ff0a1f9f0)
133+ Expected UTF-32BE: (00005fc900008ff000008aee000073ee00008f2f0002827700006b16000071b300007c6d)
134+ Actually got:      (00005fc900008ff000008aee000073ee00008f2f0000002500006b16000071b300007c6d)
135+ mb_convert_encoding incremented illegal_chars on valid EUC-JP-2004 string: (f4c6bdd2bbf0e0e3bdb4fcc9ddb48ff0a1f9f0) when converting to UTF-32BE
     EUC-JP-2004 verification and conversion works for all valid characters
137+ mb_convert_encoding not working on EUC-JP-2004 input: (fc83f6a9)
138+ Expected UTF-32BE: "%�7?" (000000250002373f)
139+ Actually got:      "%%" (0000002500000025)
140+ mb_convert_encoding not working on EUC-JP-2004 input: (f22efcd1)
141+ Expected UTF-32BE: (00000025000283cd)
142+ Actually got:      "%%" (0000002500000025)
143+ mb_convert_encoding not working on EUC-JP-2004 input: (f21ef9c7)
144+ Expected UTF-32BE: "%�Wq" (0000002500025771)
145+ Actually got:      "%%" (0000002500000025)
146+ mb_convert_encoding not working on EUC-JP-2004 input: (eefff9ee)
147+ Expected UTF-32BE: (0000002500025da1)
148+ Actually got:      "%%" (0000002500000025)
149+ mb_convert_encoding not working on EUC-JP-2004 input: (ec81f6e0)
150+ Expected UTF-32BE: (0000002500023cfe)
151+ Actually got:      "%%" (0000002500000025)
152+ mb_convert_encoding not working on EUC-JP-2004 input: (ea3ef8fe)
153+ Expected UTF-32BE: (000000250002548e)
154+ Actually got:      "%%" (0000002500000025)
155+ mb_convert_encoding not working on EUC-JP-2004 input: (e817fadd)
156+ Expected UTF-32BE: (0000002500026aff)
157+ Actually got:      "%%" (0000002500000025)
158+ mb_convert_encoding not working on EUC-JP-2004 input: (e676fbb3)
159+ Expected UTF-32BE: "%�n@" (0000002500026e40)
160+ Actually got:      "%%" (0000002500000025)
161+ mb_convert_encoding not working on EUC-JP-2004 input: (e293f9a9)
162+ Expected UTF-32BE: "%�U�" (000000250002550e)
163+ Actually got:      "%%" (0000002500000025)
164+ mb_convert_encoding not working on EUC-JP-2004 input: (e068cfe3)
165+ Expected UTF-32BE: (00000025000216b4)
166+ Actually got:      "%%" (0000002500000025)
167+ mb_convert_encoding not working on EUC-JP-2004 input: (dd23f5f2)
168+ Expected UTF-32BE: (00000025000235c4)
169+ Actually got:      "%%" (0000002500000025)
170+ mb_convert_encoding not working on EUC-JP-2004 input: (dc77cfd4)
171+ Expected UTF-32BE: (0000002500020b9f)
172+ Actually got:      "%%" (0000002500000025)
173+ mb_convert_encoding not working on EUC-JP-2004 input: (db89fcc9)
174+ Expected UTF-32BE: (0000002500028277)
175+ Actually got:      "%%" (0000002500000025)
176+ mb_convert_encoding not working on EUC-JP-2004 input: (db16f9d4)
177+ Expected UTF-32BE: (00000025000259c4)
178+ Actually got:      "%%" (0000002500000025)
179+ mb_convert_encoding not working on EUC-JP-2004 input: (d662f7ec)
180+ Expected UTF-32BE: (00000025000247f1)
181+ Actually got:      "%%" (0000002500000025)
182+ mb_convert_encoding not working on EUC-JP-2004 input: (d65dfbc9)
183+ Expected UTF-32BE: (00000025000270f4)
184+ Actually got:      "%%" (0000002500000025)
185+ mb_convert_encoding not working on EUC-JP-2004 input: (d309fbec)
186+ Expected UTF-32BE: (0000002500027684)
187+ Actually got:      "%%" (0000002500000025)
188+ mb_convert_encoding not working on EUC-JP-2004 input: (cd8baffb)
189+ Expected UTF-32BE: (00000025000218bd)
190+ Actually got:      "%%" (0000002500000025)
191+ mb_convert_encoding not working on EUC-JP-2004 input: (cc4caea2)
192+ Expected UTF-32BE: "%��" (000000250002000b)
193+ Actually got:      "%%" (0000002500000025)
194+ mb_convert_encoding not working on EUC-JP-2004 input: (ca0dafcc)
195+ Expected UTF-32BE: "%��
196+ Actually got:      "%%" (0000002500000025)
197+ mb_convert_encoding not working on EUC-JP-2004 input: (c77bfee6)
198+ Expected UTF-32BE: (000000250002a190)
199+ Actually got:      "%%" (0000002500000025)
200+ mb_convert_encoding not working on EUC-JP-2004 input: (c57bafc2)
201+ Expected UTF-32BE: "%��=" (000000250002123d)
202+ Actually got:      "%%" (0000002500000025)
203+ mb_convert_encoding not working on EUC-JP-2004 input: (c303afe0)
204+ Expected UTF-32BE: "%��n" (000000250002146e)
205+ Actually got:      "%%" (0000002500000025)
206+ mb_convert_encoding not working on EUC-JP-2004 input: (c009f6b2)
207+ Expected UTF-32BE: "%�7c" (0000002500023763)
208+ Actually got:      "%%" (0000002500000025)
209+ mb_convert_encoding not working on EUC-JP-2004 input: (be49cfee)
210+ Expected UTF-32BE: "%��4" (0000002500021e34)
211+ Actually got:      "%%" (0000002500000025)
212+ mb_convert_encoding not working on EUC-JP-2004 input: (bc36f5ba)
213+ Expected UTF-32BE: (00000025000231c4)
214+ Actually got:      "%%" (0000002500000025)
215+ mb_convert_encoding not working on EUC-JP-2004 input: (ad0aafc2)
216+ Expected UTF-32BE: "%��=" (000000250002123d)
217+ Actually got:      "%%" (0000002500000025)
218+ mb_convert_encoding not working on EUC-JP-2004 input: (ac1bfadd)
219+ Expected UTF-32BE: (0000002500026aff)
220+ Actually got:      "%%" (0000002500000025)
221+ mb_convert_encoding not working on EUC-JP-2004 input: (a569f5ba)
222+ Expected UTF-32BE: (00000025000231c4)
223+ Actually got:      "%%" (0000002500000025)
224+ mb_convert_encoding not working on EUC-JP-2004 input: (a44bafe0)
225+ Expected UTF-32BE: "%��n" (000000250002146e)
226+ Actually got:      "%%" (0000002500000025)
227+ mb_convert_encoding not working on EUC-JP-2004 input: (a218cfee)
228+ Expected UTF-32BE: "%��4" (0000002500021e34)
229+ Actually got:      "%%" (0000002500000025)
230+ mb_convert_encoding not working on EUC-JP-2004 input: (8ff97ffcc9)
231+ Expected UTF-32BE: (0000002500028277)
232+ Actually got:      "%%" (0000002500000025)
233+ mb_convert_encoding not working on EUC-JP-2004 input: (8ff66af5f2)
234+ Expected UTF-32BE: (00000025000235c4)
235+ Actually got:      "%%" (0000002500000025)
236+ mb_convert_encoding not working on EUC-JP-2004 input: (8ff22df9a9)
237+ Expected UTF-32BE: "%�U�" (000000250002550e)
238+ Actually got:      "%%" (0000002500000025)
239+ mb_convert_encoding not working on EUC-JP-2004 input: (8ff217affb)
240+ Expected UTF-32BE: (00000025000218bd)
241+ Actually got:      "%%" (0000002500000025)
242+ mb_convert_encoding not working on EUC-JP-2004 input: (8fef0cfbb3)
243+ Expected UTF-32BE: "%�n@" (0000002500026e40)
244+ Actually got:      "%%" (0000002500000025)
245+ mb_convert_encoding not working on EUC-JP-2004 input: (8fee58f8fe)
246+ Expected UTF-32BE: (000000250002548e)
247+ Actually got:      "%%" (0000002500000025)
248+ mb_convert_encoding not working on EUC-JP-2004 input: (8faf43f7ec)
249+ Expected UTF-32BE: (00000025000247f1)
250+ Actually got:      "%%" (0000002500000025)
251+ mb_convert_encoding not working on EUC-JP-2004 input: (8fa879f9c7)
252+ Expected UTF-32BE: "%�Wq" (0000002500025771)
253+ Actually got:      "%%" (0000002500000025)
254+ mb_convert_encoding not working on EUC-JP-2004 input: (8fa32efbc9)
255+ Expected UTF-32BE: (00000025000270f4)
256+ Actually got:      "%%" (0000002500000025)
257+ mb_convert_encoding not working on EUC-JP-2004 input: (8e71fbec)
258+ Expected UTF-32BE: (0000002500027684)
259+ Actually got:      "%%" (0000002500000025)
260+ mb_convert_encoding not working on EUC-JP-2004 input: (8e6bcfe3)
261+ Expected UTF-32BE: (00000025000216b4)
262+ Actually got:      "%%" (0000002500000025)

PHP Version

PHP 8.4+

Operating System

macOS 15.7.2, Apple clang version 16.0.0 (clang-1600.0.26.3)

@alexdowad

/cc @alexdowad, @youkidearitai. I don't know if either of you have a mac to reproduce this issue? I unfortunately don't. I can try setting up a Hackintosh VM (though I failed at this last time I tried).

I have a Mac and can have a look.

@iluuu1994

@iluuu1994 I built php-src on my Mac (Apple M4, running Sequoia 15.6.1), ran the tests with make test, and the failure which occurred in CI did not happen.
(Yes, I made sure that the mbstring tests are running.)
🤔

@iluuu1994

Looking at the test failure excerpt which @iluuu1994 kindly provided, it seems that every failure is essentially:
A EUC-JP-2004 input which should have been converted to a codepoint outside the BMP (>U+20000), is converted to an ASCII percent sign (%) instead.
Let me think about that and see if we can get any clue where things are going wrong...

I try reproduce, but I can't also reproduced.
macOS is 15.7.2（24G325）.

@iluuu1994

Randomly picking out one failed case for further analysis:

245+ mb_convert_encoding not working on EUC-JP-2004 input: (8fee58f8fe)
246+ Expected UTF-32BE: (000000250002548e)
247+ Actually got:      "%%" (0000002500000025)

0x8F EE 58 is junk. 0x8F is valid, the following 0xEE is valid, but 0x58 is not a legal EUC-JP-2004 byte coming after 0x8F EE, so mbstring will output an error marker (in this case, "%") and restart a new character from the next byte.

Then, 0xF8 FE is valid; it maps to U+2548E, which is the Kanji 𥒎.

If mbstring had read 0x8F EE, then (wrongly) immediately output an error marker and restarted, 0x58 would be accepted as a valid ASCII byte and would be output as U+0058. Obviously, that didn't happen. So I believe that mbstring did read 0x8F EE 58 as expected, output an error marker as expected, then something went wrong here:

/* Conversion for CJK Unified Ideographs ext.B (U+2XXXX) */
if (!w) {
	int k = mbfl_bisec_srch2(w1, jisx0213_jis_u5_key, jisx0213_u5_tbl_len);
	if (k >= 0) {
		w = jisx0213_jis_u5_tbl[k] + 0x20000;
	}
}

*out++ = w ? w : MBFL_BAD_INPUT;

(Lines 9236-9244, mbfilter_cjk.c)

That does a binary search in a table to find the conversion for codepoints >U+20000. If it doesn't find the desired character, it outputs an error marker (which is what we see).

This is the table where the binary search should occur:

static const unsigned short jisx0213_jis_u5_key[] = {
0x04C7,0x0545,0x054F,0x0563,0x057E,0x1117,0x1126,0x1131,
0x1EF1,0x1F29,0x1F3E,0x1F47,0x1F75,0x1FDF,0x204F,0x2058,
0x2076,0x2083,0x209D,0x20EA,0x211E,0x2134,0x2157,0x2192,
0x219A,0x226B,0x2284,0x228E,0x2291,0x2299,0x22A9,0x22D3,
0x22DA,0x22DC,0x22F1,0x22F8,0x22F9,0x22FB,0x22FC,0x2300,
0x2301,0x2306,0x2309,0x230B,0x230C,0x231C,0x2327,0x232D,
0x2341,0x2344,0x2346,0x2350,0x2351,0x2357,0x235E,0x2360,
0x2369,0x2371,0x2372,0x2378,0x237B,0x2396,0x23A7,0x23AE,
0x23AF,0x23B7,0x23BA,0x23D6,0x23D9,0x23DB,0x23E0,0x23E7,
0x23E8,0x23EF,0x23F1,0x23F2,0x2400,0x240D,0x2419,0x241F,
0x2422,0x2430,0x2431,0x2459,0x245B,0x2464,0x2469,0x2489,
0x249E,0x24A6,0x24AB,0x24BB,0x24C0,0x24C1,0x24C9,0x24CB,
0x24CC,0x24D0,0x24ED,0x2514,0x2518,0x2519,0x252F,0x2531,
0x2532,0x2537,0x2538,0x2539,0x253C,0x253E,0x254A,0x254B,
0x254C,0x2550,0x256C,0x256D,0x257D,0x2592,0x2593,0x2595,
0x2596,0x25A1,0x25AC,0x25B4,0x25BC,0x25BD,0x25C3,0x25C8,
0x25D4,0x25E5,0x25FA,0x260D,0x260F,0x2610,0x2611,0x2641,
0x2656,0x265C,0x2670,0x2673,0x268F,0x26A0,0x26A6,0x26C0,
0x26E8,0x26F9,0x26FB,0x2700,0x270F,0x2728,0x272C,0x2731,
0x2734,0x2740,0x2742,0x2745,0x274A,0x274C,0x274D,0x2751,
0x2755,0x2766,0x2771,0x2784,0x279E,0x279F,0x27B9,0x27C4,
0x27C5,0x27C7,0x27D9,0x27E4,0x27E5,0x27FA,0x27FB,0x27FC,
0x27FE,0x2802,0x2804,0x2807,0x2809,0x280C,0x2813,0x2814,
0x2819,0x281A,0x2822,0x2827,0x2834,0x284E,0x2850,0x2857,
0x285A,0x285E,0x2878,0x287D,0x2889,0x2899,0x289B,0x289D,
0x28A0,0x28A2,0x28A6,0x28AD,0x28B3,0x28B6,0x28E5,0x28EF,
0x28FE,0x2916,0x291F,0x2920,0x2921,0x2932,0x2935,0x2963,
0x2964,0x296A,0x296D,0x2972,0x2986,0x2987,0x2989,0x2991,
0x2999,0x299B,0x299F,0x29B3,0x29C0,0x29D4,0x29D6,0x29D7,
0x29E0,0x29EA,0x29ED,0x29F4,0x29FD,0x2A03,0x2A14,0x2A19,
0x2A21,0x2A26,0x2A35,0x2A39,0x2A3A,0x2A45,0x2A48,0x2A68,
0x2A69,0x2A70,0x2A7E,0x2A7F,0x2A8A,0x2A8B,0x2A97,0x2A98,
0x2AA3,0x2AA4,0x2AAD,0x2AAE,0x2AB4,0x2AB5,0x2AC5,0x2AC6,
0x2ACE,0x2AD1,0x2AD3,0x2AD4,0x2AD8,0x2ADC,0x2ADE,0x2AE0,
0x2AE8,0x2AF7,0x2AF8,0x2B0D,0x2B17,0x2B21,0x2B2B,0x2B2E,
0x2B32,0x2B4B,0x2B5F,0x2B7E,0x2B7F,0x2B8C,0x2B8E,0x2B90,
0x2B9A,0x2BA0,0x2BA3,0x2BB6,0x2BBA,0x2BBC,0x2BC3,0x2BC6,
0x2BE4,0x2BE9,0x2BEB,0x2BFF,0x2C01,0x2C03,0x2C07
};

In the failing CI run which @iluuu1994 kindly shared, many EUC-JP-2004 conversion operations involving codepoints >U+20000 all failed. That implies that the problem was not caused by something like a corrupted bit or byte somewhere in this table. Otherwise, we would expect that only a couple of test cases would fail.

The function mbfl_bisec_srch2 is used for many Japanese and Chinese text encodings. But we are only seeing test failures for EUC-JP-2004. That implies that the problem was not caused by a miscompilation of mbfl_bisec_srch2.

Further, the table jisx0213_jis_u5_key is not just used for EUC-JP-2004, but also for SJIS-2004 and ISO-2022-JP-2004. But we don't see any test failures for those encodings. So again, it doesn't look like the problem is caused by corruption of jisx0213_jis_u5_key.

If anything, what I am seeing seems consistent with a miscompilation or corruption of this little snippet of code itself:

if (!w) {
	int k = mbfl_bisec_srch2(w1, jisx0213_jis_u5_key, jisx0213_u5_tbl_len);
	if (k >= 0) {
		w = jisx0213_jis_u5_tbl[k] + 0x20000;
	}
}

Any thoughts from others? @iluuu1994 @youkidearitai @cmb69 @ndossche

Thank you for your analysis! I'll try to pull the quoted code into an unoptimized function to see if it changes anything.

@alexdowad

And voilà:

iluuu1994#168

This indeed seems to be the culprit. Very nice deduction by @alexdowad.

@alexdowad @youkidearitai Could you tell me which version of Clang you've tried this with? It would be optimal if we could pin-point the issue and even share a diff of the assembly to report an issue, but this is very hard (or rather extremely tedious) to do via CI.

7:02 ~/Programming/php-src % clang -v
Apple clang version 17.0.0 (clang-1700.0.13.5)
Target: arm64-apple-darwin24.6.0
Thread model: posix
InstalledDir: /Library/Developer/CommandLineTools/usr/bin

In the failing CI test run, I see this:

Apple clang version 16.0.0 (clang-1600.0.26.3)
Target: x86_64-apple-darwin24.6.0
Thread model: posix
InstalledDir: /Applications/Xcode_16.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin

I'll try installing Clang 16 on my Mac and see if I can replicate the problem that way.

@alexdowad

@alexdowad Thank you. Good to know. CI runs with clang-1600.0.26.3, so possibly this issue is already fixed.

I'll try installing Clang 16 on my Mac and see if I can replicate the problem that way.

Thank you!

Just recompiling now. I did brew install llvm@16, and it installed Clang 16.0.6... not exactly the same version which the CI failures are happening for 😢
Anyways, let's see what will happen.

@alexdowad

@alexdowad Heads-up: Confusingly, Apple Clang and Clang/LLVM versions do not match. See:

https://en.wikipedia.org/wiki/Xcode#Xcode_15.0_-_16.x_(since_visionOS_support)_2

Apple Clang 16.0.0 is actually Clang/LLVM 17.0.6.

And of course now the Xcode version also diverges from Apple Clang, making things even more confusing...

Also, it's possible using LLVM Clang will not actually reproduce the issue. I tried Clang on Linux without success. Possibly it might be reproducible with Xcode 16.x.

@iluuu1994

@iluuu1994 Hmm.
At any rate, after recompiling with the version of Clang which I got with brew install llvm@16, I tried running the test in question and didn't see any failure.

@iluuu1994

@iluuu1994 Crazy idea here... maybe temporarily add a CI step to dump the disassembly of the problem function?
I'd be happy to analyze the assembler code and see if I can find any problem with it.

I realized... We're actually installing Xcode 26, but the old version sticks around, which is the version ls picks up and sets as the default. So, we can install Xcode 26, and the issue goes away there. Rather than investing more time, I think it's easiest if we just do that.

The line that should fix things:

https://github.com/iluuu1994/php-src/pull/168/changes#diff-75c23100d1f02f80e48f89431728570b782f2e4440bce0a6acd41d1fe962ce08R11

This will break in 74 years because macOS doesn't have natural sort (-V), but somebody else can fix that. 😛

Edit: 6cdb8bf, let's see if that fixes nightly tomorrow.

@iluuu1994

Awesome find, @iluuu1994. 👍🏻👍🏻👍🏻

Some observations in case in the hope it might help any.
The sort command on macOS absolutely does have -V aka --version-sort.
That macos-update-clang/action.yml script looks very strange to me. It is using the software_update command to download & install the latest of the Command Line Tools for Xcode, and if any found -- and only then --, executes xcode-select with a path that looks like /Applications/Xcode_*.app. The Command Line Tools for Xcode package actually installs into /Library/Developer/CommandLineTools and does not touch the /Applications folder or install the Xcode app at all. In effect it is actively ignoring and disregarding whatever new version of the command line tools package it might have installed, and is resetting the developer tools to point elsewhere. If xcode-select -s is to be called with any path after a successful installation of the command line tools package, it should be /Library/Developer/CommandLineTools — however I would assume the mere installation of the package would already do that automatically. This script is only overriding that right away, turning the whole point of installing the new package pointless.

@iluuu1994

@iluuu1994 For clarity... has this issue with our CI been resolved? If so, I guess we could close this issue?

Yes, thanks for the heads-up. 🙂

Navigation Menu

Repository navigation

eucjp_2004_encoding.phpt fails on macOS in release builds #20686

Description

PHP Version

Operating System

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

eucjp_2004_encoding.phpt fails on macOS in release builds #20686

Description

Description

PHP Version

Operating System

Activity

iluuu1994 commented on Dec 11, 2025

alexdowad commented on Dec 11, 2025

alexdowad commented on Dec 12, 2025

alexdowad commented on Dec 12, 2025

youkidearitai commented on Dec 12, 2025

alexdowad commented on Dec 12, 2025

iluuu1994 commented on Dec 12, 2025

iluuu1994 commented on Dec 12, 2025

alexdowad commented on Dec 12, 2025

alexdowad commented on Dec 12, 2025

iluuu1994 commented on Dec 12, 2025

alexdowad commented on Dec 12, 2025

iluuu1994 commented on Dec 12, 2025

iluuu1994 commented on Dec 12, 2025

alexdowad commented on Dec 12, 2025

alexdowad commented on Dec 12, 2025

iluuu1994 commented on Dec 13, 2025

alexdowad commented on Dec 13, 2025

fds242 commented on Dec 17, 2025

alexdowad commented on Nov 8, 2026

iluuu1994 commented on Nov 9, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions