Inconsistent dump results of jq #2367
Comments
|
Reproduced with debian 12 and gcc: root@6eeb8d61a151:/Users/wader/src/jq# gcc --version
gcc (Debian 12.2.0-14) 12.2.0
Copyright (C) 2022 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
root@6eeb8d61a151:/Users/wader/src/jq# echo '{"a": 1E999999999}' | ./jq
{
"a": 1E+999999999
}
root@6eeb8d61a151:/Users/wader/src/jq# echo '{"a": 1E9999999999}' | ./jq
{
"a": ((double)1.79769313486231570814527423731704357e+308L)
} |
|
Seems to boil down to this difference: root@6eeb8d61a151:/Users/wader/src/jq# echo -e '#include <float.h>\n#define STR(x) #x\n#define XSTR(x) STR(x)\nXSTR(DBL_MAX) DBL_MAX __DBL_MAX__' | clang -E -
# 1 "<stdin>"
# 1 "<built-in>" 1
# 1 "<built-in>" 3
# 361 "<built-in>" 3
# 1 "<command line>" 1
# 1 "<built-in>" 2
# 1 "<stdin>" 2
# 1 "/usr/lib/llvm-14/lib/clang/14.0.6/include/float.h" 1 3
# 2 "<stdin>" 2
"1.7976931348623157e+308" 1.7976931348623157e+308 1.7976931348623157e+308
root@6eeb8d61a151:/Users/wader/src/jq# echo -e '#include <float.h>\n#define STR(x) #x\n#define XSTR(x) STR(x)\nXSTR(DBL_MAX) DBL_MAX __DBL_MAX__' | cpp
# 0 "<stdin>"
# 0 "<built-in>"
# 0 "<command-line>"
# 1 "/usr/include/stdc-predef.h" 1 3 4
# 0 "<command-line>" 2
# 1 "<stdin>"
# 1 "/usr/lib/gcc/x86_64-linux-gnu/12/include/float.h" 1 3 4
# 2 "<stdin>" 2
"((double)1.79769313486231570814527423731704357e+308L)" ((double)1.79769313486231570814527423731704357e+308L) ((double)1.79769313486231570814527423731704357e+308L)
|
|
I wonder what the backwards compatibiltiy is about at https://github.com/jqlang/jq/blob/master/src/jv.c#L660-L667 to preserve how Otherwise i guess we could just skip the Other alternatives could be to runtime string format the constants at startup or maybe define our own min/max constants somehow? |
|
Removing these statements outputs |
|
That's not good. So that is the case when the decNumber code thinks it's too large even |
Hi there,
I found that for large floating numbers, when compiling with gcc and clang,
jqwould emit different outputs.For example, for
echo '{"a": 1E9999999999}' | jq .,gcc compiled
jqwould emit:while clang compiled
jqwould emit:I think the cause is probably different interpretation of
DBL_MAX/DBL_MINin gcc and clang. But to ensure the correctness, robustness, and consistency ofjq, it's important to keep programs' behaviour consistent across main-stream compilers.The text was updated successfully, but these errors were encountered: