Use -1 "precision" in gen_stub.php #8734
Conversation
| @@ -6,7 +6,7 @@ | |||
| * @var float | |||
| * @cname M_E | |||
| */ | |||
| const M_E = 2.7182818284590452354; | |||
| const M_E = 2.718281828459045; | |||
There was a problem hiding this comment.
There was a problem hiding this comment.
why is M_E declared here and no other math constants?
There was a problem hiding this comment.
Oh, indeed, that looks fishy. We use the definitions provided by libc, if available, and only define the macros ourselves otherwise. There is nothing special about M_E compared to the other constants (like M_PI).
However, given that we already go to great lengths to have portable floating point arithmetic, it may not be the best idea to rely on definitions provided by libc.
| @@ -3518,7 +3518,7 @@ static const zend_function_entry class_AssertionError_methods[] = { | |||
| static void register_basic_functions_symbols(int module_number) | |||
| { | |||
| REGISTER_DOUBLE_CONSTANT("M_E", M_E, CONST_CS | CONST_PERSISTENT); | |||
| ZEND_ASSERT(M_E == 2.7182818284590451); | |||
| ZEND_ASSERT(M_E == 2.718281828459045); | |||
There was a problem hiding this comment.
needed as compared as string here, https://3v4l.org/9C9LV
| @@ -1,7 +1,7 @@ | |||
| --TEST-- | |||
| Math constants | |||
| --INI-- | |||
| precision=14 | |||
| precision=-1 | |||
There was a problem hiding this comment.
we want to assert the full precision, double is used for both x86/x64 build
| @@ -1,7 +1,5 @@ | |||
| --TEST-- | |||
| Test for pre-defined math constants | |||
| --INI-- | |||
| precision=14 | |||
There was a problem hiding this comment.
var_dump uses (correctly) always the full/-1 precision
| M_SQRT1_2 : 0.707106[0-9]* | ||
| M_SQRT3 : 1.732050[0-9]* | ||
| --EXPECT-- | ||
| M_E : 2.718281828459045 |
There was a problem hiding this comment.
on all platforms, the output should be strictly the same
|
@bwoebi, @kocsismate, what do you think about this PR? |
|
@bwoebi @kocsismate @cmb69 this PR sits here for a long time but is finished, can you please review? |
|
This looks reasonable, but needs a rebase. |
|
rebased |
|
Thank you! |
https://3v4l.org/FkrAj - only
-1results in the shortest 64-bit IEEE 754 floating-point representation possible,17can print more unneeded digitsalso improve two math constants tests by asserting the exact values