summaryrefslogtreecommitdiff
path: root/toke.c
diff options
context:
space:
mode:
authorJarkko Hietaniemi <jhi@iki.fi>2000-05-30 22:53:37 +0000
committerJarkko Hietaniemi <jhi@iki.fi>2000-05-30 22:53:37 +0000
commitd6c1400008ffe28d0a765e9fe60d23dbeab89dda (patch)
tree8ebf4512d2d54f3e1821a457a5d9e82b016966a9 /toke.c
parent5b0d9cefa224486d66580e5164324f002f3989ab (diff)
downloadperl-d6c1400008ffe28d0a765e9fe60d23dbeab89dda.tar.gz
Tweak NV_PRESERVES_UV*, vms/subconfigure.com left untouched.
p4raw-id: //depot/cfgperl@6168
Diffstat (limited to 'toke.c')
-rw-r--r--toke.c22
1 files changed, 13 insertions, 9 deletions
diff --git a/toke.c b/toke.c
index 75cab919cd..99c993e8fa 100644
--- a/toke.c
+++ b/toke.c
@@ -7025,7 +7025,9 @@ Perl_scan_num(pTHX_ char *start)
Strtol() and Strtoul() are used above.
[1] XXX Configure test needed to check for atol()
- (and atoll() overflow behaviour) XXX --jhi
+ (and atoll()) overflow behaviour XXX
+
+ --jhi
We need to do this the hard way. */
@@ -7039,14 +7041,16 @@ Perl_scan_num(pTHX_ char *start)
don't need to do the conversion at all.
[1] Note that this is lossy if our NVs cannot preserve our
- UVs. There is a metaconfig define, NV_PRESERVES_UV, but we
- really do hope all such platforms have strtou?ll? to do a
- lossless IV/UV conversion.
- XXX Configure test needed to check how many UV bits
- do our NVs preserve, really (the current test checks
- for the roundtrip of ~0) XXX --jhi
- Maybe do some tricks with DBL_MANT_DIG and LDBL_MANT_DIG,
- and DBL_DIG, LDBL_DIG (this is already available as NV_DIG)?
+ UVs. There are metaconfig defines NV_PRESERVES_UV (a boolean)
+ and NV_PRESERVES_UV_BITS (a number), but in general we really
+ do hope all such potentially lossy platforms have strtou?ll?
+ to do a lossless IV/UV conversion.
+
+ Maybe could do some tricks with DBL_DIG, LDBL_DIG and
+ DBL_MANT_DIG and LDBL_MANT_DIG (these are already available
+ as NV_DIG and NV_MANT_DIG)?
+
+ --jhi
*/
{
UV uv = U_V(nv);