You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Compare test program output with the other Haxe targets #150
Run the same test programs on hxcpp and on HashLink, eval and JS and diff the output, to find semantic differences systematically instead of one report at a time (see #97 to #135).
Done in 47525836b503f756b9e5a7cbd7f5b45d128bb2e0 and abb37f30d0fc4c5b6e3b61339e9bd2a909df342c. test/targets runs the same programs on C++, cppia (interpreted and with the JIT), JS, eval, Neko, Python and, where CI has them, HashLink, JVM and PHP, and diffs each with C++. A cppia line that differs fails the job; the other targets are reported, each differing line marked with its reason from known-differences.txt or flagged as not explained.
What it has found so far: the cppia JIT boxing Array Bool results as Int (fixed in a7273a867aa6002e1c858cb225aef54b3c941274, #260). Neko matches C++ on every line, and cppia does too, in both modes. The eval and Python differences are known target differences, apart from two that are eval's own and one in haxe.crypto.Sha256, which gives the String's length rather than its UTF-8 byte count as the message length.
Activity
Box Array Bool results as Bool in the cppia jit
Box Array Bool results as Bool in the cppia jit
Box Array Bool results as Bool in the cppia jit
Compare program output across Haxe targets
Explain eval's and Python's differences, and let Python run
Compare program output across Haxe targets
Explain eval's and Python's differences, and let Python run
dimensionscape commentedon Oct 23, 2026
Done in
47525836b503f756b9e5a7cbd7f5b45d128bb2e0andabb37f30d0fc4c5b6e3b61339e9bd2a909df342c. test/targets runs the same programs on C++, cppia (interpreted and with the JIT), JS, eval, Neko, Python and, where CI has them, HashLink, JVM and PHP, and diffs each with C++. A cppia line that differs fails the job; the other targets are reported, each differing line marked with its reason from known-differences.txt or flagged as not explained.What it has found so far: the cppia JIT boxing Array Bool results as Int (fixed in
a7273a867aa6002e1c858cb225aef54b3c941274, #260). Neko matches C++ on every line, and cppia does too, in both modes. The eval and Python differences are known target differences, apart from two that are eval's own and one in haxe.crypto.Sha256, which gives the String's length rather than its UTF-8 byte count as the message length.