Pointer cast to integer of different size in hv_func.h #12938
Comments
From rmbarker.cpan@btinternet.comCreated by rmbarker.cpan@btinternet.comWhen compiling perl, I get many warnings about the same line in The attached patch, also changes the type of the variable on that Perl Info |
From rmbarker.cpan@btinternet.com0001-avoid-cast-from-pointer-to-integer-of-different-size.patchFrom a3801c134954362fb394b6c1ffbefe0d1c045d4e Mon Sep 17 00:00:00 2001
From: Robin Barker <rmbarker@cpan.org>
Date: Wed, 1 May 2013 18:39:26 +0100
Subject: [PATCH 1/2] avoid cast from pointer to integer of different size
---
hv_func.h | 4 ++--
1 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/hv_func.h b/hv_func.h
index 0a66356..9dfe158 100644
--- a/hv_func.h
+++ b/hv_func.h
@@ -392,8 +392,8 @@ S_perl_hash_murmur3(const unsigned char * const seed, const unsigned char *ptr,
/* This CPU does not handle unaligned word access */
/* Consume enough so that the next data byte is word aligned */
- int i = -(IV)ptr & 3;
- if(i && (STRLEN)i <= len) {
+ STRLEN i = -PTR2IV(ptr) & 3;
+ if(i && i <= len) {
MURMUR_DOBYTES(i, h1, carry, bytes_in_carry, ptr, len);
}
--
1.7.5.4
|
From @demerphqSorry about that, probably my bad porting skills at fault. I will Yves On 2 May 2013 12:14, Robin Barker <perlbug-followup@perl.org> wrote:
-- |
|
The RT System itself - Status changed from 'new' to 'open' |
From @nwc10On Thu, May 02, 2013 at 12:23:18PM +0200, demerphq wrote:
It needs testing on Win64 first. Nicholas Clark |
From @demerphqOn 2 May 2013 12:34, Nicholas Clark <nick@ccl4.org> wrote:
Any reason Win64 is special in this regard? Yves -- |
From @nwc10On Thu, May 02, 2013 at 12:46:30PM +0200, demerphq wrote:
Because the warning is not from your code. It's from a change made to your https://rt-archive.perl.org/perl5/Ticket/Display.html?id=117687 So changing it has the potential to break their build. I'd spotted this warning a few days ago and had pushed a branch recently Also, I don't understand why it was a build *failure* on Win64, and not just Nicholas Clark |
From @demerphqOn 2 May 2013 13:07, Nicholas Clark <nick@ccl4.org> wrote:
Oh. So maybe we should leave it one of our people who can easily test Yves -- |
From @steve-m-haydemerphq wrote on 2013-05-02:
I'll give it whirl shortly... :-) |
From @sisyphus-----Original Message-----
It was I that created that ticket. I've tried Robin's patch on 5.17.11 - it's fine for me on Win64. (I haven't
It *is* strange. I also use 4.7.0 for my 32-bit builds. There was no problem there ... but, Cheers, |
From @nwc10On Thu, May 02, 2013 at 10:33:12PM +1000, sisyphus1@optusnet.com.au wrote:
Cool, thanks. I hadn't yet asked you, because I was going to verify that the
Well, they are by default. The warning is cropping up on *nix when one Also, Robin's attached *patch* has the 1 line commit message which doesn't reference the RT number (it can't - RT allocated the ticket commit 275227021431a7dbfb1c18e88a68a35ce0851161 Use PTR2IV() instead of casting a pointer to IV, to avoid warnings. Inline Patchdiff --git a/hv_func.h b/hv_func.h
index 0a66356..9dfe158 100644
--- a/hv_func.h
+++ b/hv_func.h
@@ -392,8 +392,8 @@ S_perl_hash_murmur3(const unsigned char * const seed, const
/* This CPU does not handle unaligned word access */
/* Consume enough so that the next data byte is word aligned */
- int i = -(IV)ptr & 3;
- if(i && (STRLEN)i <= len) {
+ STRLEN i = -PTR2IV(ptr) & 3;
+ if(i && i <= len) {
MURMUR_DOBYTES(i, h1, carry, bytes_in_carry, ptr, len);
}
Nicholas Clark |
From @steve-m-hayNicholas Clark wrote on 2013-05-02:
It works fine for me too, using MinGW 4.7.0 (32-bit native compiler), [...]
warnings.
pointers.
everyone
macro
I've also tested with VC++ 2010 32-bit and 64-bit native compilers. No hv_func.h(397) : warning C4267: 'initializing' : conversion from However, it already has several other warnings from that file: hv_func.h(264) : warning C4267: 'initializing' : conversion from and generally gazillions of other warnings throughout the build which I haven't tried a 32-bit build with 64-bit-ints yet, but I will do later PS. I also had Robin's other patch (RT #117837) in place throughout all |
From @steve-m-haySteve Hay wrote on 2013-05-02:
I've done 64-bit-int builds with the 32-bit MinGW and VC++ 2010 |
From @nwc10On Thu, May 02, 2013 at 06:05:58PM +0100, Steve Hay wrote:
Cool. The smoke tests all look clean. I think that we should merge this fix to blead before v5.18.0 Whilst it's only a compiler warning fix, not a build breaker, it's clear Hence I'll mark this ticket as a blocker and Ricardo can unmark it if he Nicholas Clark |
From @rjbs* Nicholas Clark <nick@ccl4.org> [2013-05-05T15:27:48]
Yes. -- |
|
@nwc10 - Status changed from 'open' to 'resolved' |
Migrated from rt.perl.org#117835 (status was 'resolved')
Searchable as RT117835$
The text was updated successfully, but these errors were encountered: