Comparing 9d29b8329b..efa3896fa4 - httrack - Synthesist

rms/httrack

mirror of https://github.com/xroche/httrack.git synced 2026-07-03 07:33:49 +03:00

Author	SHA1	Message	Date
Xavier Roche	efa3896fa4	tests: cover the %xx-encoded UTF-8 flush path in unescape-bounds The raw-byte cases never take the utfBufferJ = lastJ rollback branch, so a wrong flush offset there would have passed (review finding). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Signed-off-by: Xavier Roche <roche@httrack.com>	2026-07-02 22:37:55 +02:00
Xavier Roche	d2b71d878c	configure: use the dylib plugin name on Darwin libtool names the module libhtsjava.N.dylib there, so the .so.N form can never load; caught by 31_local-javaclass.test on the macOS CI job (the old hardcoded .so.2 was just as dead, silently). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Signed-off-by: Xavier Roche <roche@httrack.com>	2026-07-02 22:37:55 +02:00
Xavier Roche	1249c42f1e	htsftp: assert nonzero buffer sizes, harden the userpass self-test ftp_split_userpass underflows its size-1 math on a zero size; assert the precondition now that the function is public in htsftp.h. The self-test gains a tight-size run with guard bytes and exact-content checks, which the 256-byte buffers alone could not fail on an off-by-one. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Signed-off-by: Xavier Roche <roche@httrack.com>	2026-07-02 22:28:04 +02:00
Xavier Roche	5c1ffaf3a6	Resurrect the java .class parser on modern Unix builds The plugin was dead three ways on a current Linux build: hts_plug() was compiled hidden (-fvisibility=hidden, EXTERNAL_FUNCTION expanded to nothing on ELF), hts_create_opt() dlopens libhtsjava.so.2 which no longer exists since the soname moved to .so.3, and JAVA_HEADER.magic is 'unsigned long' (8 bytes on LP64) under a 10-byte fread, so major/count came from uninitialized bytes and the 0xCAFEBABE check never matched. EXTERNAL_FUNCTION now forces default visibility on ELF, the dlopen name is derived from VERSION_INFO at configure time, and the header fields are fixed-width. 31_local-javaclass.test crawls a generated .class and asserts a resource named only in its constant pool is fetched; it fails if any of the three regresses. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Signed-off-by: Xavier Roche <roche@httrack.com>	2026-07-02 22:27:55 +02:00
Xavier Roche	f38472c697	htsencoding: bound the raw UTF-8 flush in hts_unescapeUrlSpecial The completed-sequence flush memcpy ends with a 'continue' that skips the per-byte NUL-reserve guard, so a raw multi-byte character landing at the exact end of dest let the trailing NUL write dest[max] (1-byte OOB, found by the post-#474 review pass; ASan-verified via the extended -#test=unescape-bounds). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Signed-off-by: Xavier Roche <roche@httrack.com>	2026-07-02 22:27:43 +02:00

Diff Content Not Available