[kairosci]

Fixes #24170 by reading concatenated bzip2 members in the stream wrapper and bzopen(), while preserving the exported API and trailing-data behavior.

Adds regression coverage; bzip2 tests pass in debug ZTS and ASan/UBSan builds with applicable skips.

[kairosci]

The macOS job fails during compilation because OpenSSL 4.1 deprecates ASN1_STRING_set() and ASN1_STRING_length(), and deprecation warnings are treated as errors. Tests never start. This is unrelated to the bzip2 changes and is already addressed separately by #24686.

Failing job: https://github.com/php/php-src/actions/runs/40205742173/job/121206203893

[anun333]

Quoted below: written with Claude, an LLM, as CONTRIBUTING.md asks us to mark. The builds and test runs behind it are on my machine; the commands to repeat them are under the fold.

Tested on today's master (1a72e5f2542) with this PR cherry-picked, debug builds with libbz2 1.0.8. Files made with bzip2 1.0.8, every output compared with bzcat's by length and sha256. "Stream readers" is file_get_contents('compress.bzip2://…'), bzopen($path) + bzread(), bzopen($fp) + bzread() and fread($h, 7) on the wrapper; they agreed in every row:

File	bzcat	stream readers, master	stream readers, this PR	bzdecompress(), both
one member	143,379 bytes	match	match	match
two members	215,128	the first member only	match	the first member only
fifty members, as a parallel compressor writes them	143,379	2,868	match	2,868
an empty member between two	215,128	the first member only	match	the first member only
trailing trailing junk\n	143,379	match	match	match
trailing BZh9 and random bytes	the first member, exit 2	the first member	the first member	the first member
the second member cut in half	the first member, exit 2	the first member	the first member	the first member

ext/bz2/tests passes on both (31; 34 with the three new tests), and the three new tests fail on master without the fix.

Two things I noticed:

• bzdecompress() still stops after the first member, so on these files the string API and the stream API now give different results. The UPGRADING entry names only the wrapper and bzopen(); it could say which applies to bzdecompress().
• When a later member is damaged, as in the last row, bzcat exits with "Compressed file ends unexpectedly". With this PR, bzerror() afterwards reports UNEXPECTED_EOF (master says OK), but bzread() and fread() never return false and file_get_contents() returns the first member's 143,379 bytes without a warning. A single truncated member does make bzread() return false, so a cut-off multi-member file reads as complete unless the caller checks bzerror().

How this was tested

Linux x86_64, gcc 13.3, debug builds of origin/master and of it with this PR's commit cherry-picked:

./buildconf --force
./configure --enable-debug --disable-cgi --disable-phpdbg --disable-fileinfo --enable-pcntl --enable-zend-test --enable-fpm --with-curl --enable-mbstring --disable-mbregex --with-mysqli --with-pdo-mysql --with-openssl --with-bz2
make -j4

Two texts of 143,379 and 71,749 bytes; bzip2 -c a.txt; bzip2 -c b.txt into one file, and so on for each row (the fifty-member file is split -n 50 of the first text, each part compressed); the truncated row is the first member plus the first half of the second. Each reader loops until EOF (bzread($bz, 8192), or fread($h, 7)), stopping on false. Suites: TEST_PHP_EXECUTABLE=$PWD/sapi/cli/php sapi/cli/php run-tests.php -n -q -j4 ext/bz2/tests.

[kairosci]

@ndossche As soon as you can, could you take a look at it ? Thanks