Unsuppressable warning under GCC causing build failure with -Werror (again)

[GoogleCodeExporter]

Similar to my previous defect report 
(http://code.google.com/p/pugixml/issues/detail?id=217). Not sure why this 
can't be disabled via #pragma. Using a prerelease build of GCC v4.9.0 this time 
(which is why I assume I'm only seeing this now).

Repros on both x64 and x86 release-mode build configurations for me (full 
command line with relevant flags and switches below).

Build output from one of my tests that consumes PugiXML:
==========
gcc.compile.c++ 
bin\tests\find_pattern.test\gcc-mingw-4.9.0\release\address-model-32\architectur
e-x86\debug-symbols-on\threading-multi\find_pattern.o
In file included from include\memory/hadesmem/find_pattern.hpp:21:0,
                 from tests\find_pattern.cpp:4:
C:\Code\pugixml-trunk\src/pugixml.cpp: In function 
'pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*)':
C:\Code\pugixml-trunk\src/pugixml.cpp:956:10: error: assuming signed overflow 
does not occur when changing X +- C1 cmp C2 to X cmp C1 +- C2 
[-Werror=strict-overflow]
     else if (static_cast<unsigned int>(lead - 0xD800) < 0x400 && data + 1 < end)
          ^
cc1plus.exe: all warnings being treated as errors

    "g++"  -ftemplate-depth-128 -O3 -finline-functions -Wno-inline -Wall -pedantic -Werror -g -march=i686 -mthreads -m32 -ansi -Wpedantic -Wextra -Weffc++ -Wshadow -Wconversion -Winit-self -Wmissing-include-dirs -Wstrict-aliasing -Wstrict-overflow=5 -Wno-effc++ -Wold-style-cast -Wnon-virtual-dtor -Woverloaded-virtual -Winvalid-pch -Wno-multichar -flto -std=c++1y -DASMJIT_API= -DBOOST_FILESYSTEM_NO_DEPRECATED -DBOOST_SYSTEM_NO_DEPRECATED -DBOOST_USE_WINDOWS_H -DNDEBUG -DPUGIXML_HEADER_ONLY -DPUGIXML_WCHAR_MODE -DSTRICT -DSTRICT_TYPED_ITEMIDS -DUNICODE -DWINVER=_WIN32_WINNT_VISTA -D_CRT_SECURE_NO_WARNINGS -D_SCL_SECURE_NO_WARNINGS -D_UNICODE -D_WIN32_WINNT=_WIN32_WINNT_VISTA  -I"C:\Code\asmjit-trunk\asmjit\src" -I"C:\Code\boost-trunk" -I"C:\Code\pugixml-trunk\src" -I"C:\Code\udis86-trunk" -I"include\memory" -c -o "bin\tests\find_pattern.test\gcc-mingw-4.9.0\release\address-model-32\architecture-x86\debug-symbols-on\threading-multi\find_pattern.o" "tests\find_pattern.cpp"
==========

Thanks.

Original issue reported on code.google.com by raptorfactor@raptorfactor.com on 3 Dec 2013 at 5:32

[GoogleCodeExporter]

I do not see the issue in gcc 4.8.1; I see another problem that I missed before 
because I did not verify wchar mode:

pugixml.cpp: In function ‘pugi::impl::(anonymous 
namespace)::convert_buffer(wchar_t*, unsigned char*, unsigned short*, unsigned 
int*, wchar_t const*, unsigned long, pugi::xml_encoding)’:
cc1plus: error: assuming pointer wraparound does not occur when comparing P +- 
C1 with P +- C2 [-Werror=strict-overflow]

(note the absence of a line number)

Your warning message looks dangerous though. It seems to suggest that gcc 
rewrites the code from
(unsigned)(lead - 0xd800) (cmp) 0x400
to
lead (cmp) 0x400 + 0xD800

which can't be correct. Maybe the generated code is actually correct and it's 
just the warning message that's weird. Can you post disassembly for 
decode_utf16_block in this build?

Additionally, unless I'm misreading the error message, the signed overflow 
can't really occur here (that was true for the previous issue as well).

Is this the only warning? 
Can you check if changing the type of 'lead' here:

uint16_t lead = opt_swap::value ? endian_swap(*data) : *data;

To 'unsigned int' fixes the problem?

Original comment by arseny.k...@gmail.com on 3 Dec 2013 at 7:34

[GoogleCodeExporter]

Hey, I won't be able to check until tomorrow (no compiler on this machine), but 
I think what might be happening is that "lead" is being promoted to int due to 
the subtraction of 0xd800 (which is signed!). You only cast it back to unsigned 
after the subtraction, at which point the 'overflow' could have potentially 
already occurred (can't check the surrounding code, because again I'm not on 
the machine with my code/compilers, but based on the error message that is what 
I would assume...).

I'm thinking that perhaps changing the 0xd800 to 0xd800U might solve it too, 
but I'll let you know tomorrow.

Thanks for the quick response.

Original comment by raptorfactor@raptorfactor.com on 3 Dec 2013 at 11:32

[GoogleCodeExporter]

Yes, both arguments are promoted to int; however signed overflow can't possibly 
occur here, and I expected the compiler to know that since I thought it 
performs some form of range tracking for values. Maybe it's not utilized in 
this case.

Original comment by arseny.k...@gmail.com on 3 Dec 2013 at 3:45

[GoogleCodeExporter]

Changing 'lead' to 'unsigned int' does indeed fix the problem (didn't try 
anything else due to time constraints).

I've pasted below the full disassembly for the function in Release x86. Haven't 
had time yet to look at it and figure out if it's correct or not. Probably 
won't have time until tonight or tomorrow. I expect you'll beat me to it. :)

Thanks again.

(gdb) disas 0x407f60
Dump of assembler code for function pugi::impl::utf_decoder<pugi::impl::utf16_wr
iter, pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned
 int, unsigned short*):
   0x00407f60 <+0>:     push   %ebp
   0x00407f61 <+1>:     push   %edi
   0x00407f62 <+2>:     push   %esi
   0x00407f63 <+3>:     push   %ebx
   0x00407f64 <+4>:     mov    0x14(%esp),%edx
   0x00407f68 <+8>:     mov    0x18(%esp),%ecx
   0x00407f6c <+12>:    mov    0x1c(%esp),%eax
   0x00407f70 <+16>:    lea    (%edx,%ecx,2),%esi
   0x00407f73 <+19>:    cmp    %esi,%edx
   0x00407f75 <+21>:    jae    0x407fe6 <pugi::impl::utf_decoder<pugi::impl::utf
16_writer, pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, uns
igned int, unsigned short*)+134>
   0x00407f77 <+23>:    mov    %esi,%esi
   0x00407f79 <+25>:    lea    0x0(%edi,%eiz,1),%edi
   0x00407f80 <+32>:    movzwl (%edx),%ecx
   0x00407f83 <+35>:    cmp    $0xd7ff,%cx
   0x00407f88 <+40>:    jbe    0x407ff0 <pugi::impl::utf_decoder<pugi::impl::utf
16_writer, pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, uns
igned int, unsigned short*)+144>
   0x00407f8a <+42>:    movzwl %cx,%ebx
   0x00407f8d <+45>:    sub    $0xe000,%ebx
   0x00407f93 <+51>:    cmp    $0x1fff,%ebx
   0x00407f99 <+57>:    jbe    0x407ff0 <pugi::impl::utf_decoder<pugi::impl::utf
16_writer, pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, uns
igned int, unsigned short*)+144>
   0x00407f9b <+59>:    cmp    $0xdbff,%cx
   0x00407fa0 <+64>:    ja     0x408002 <pugi::impl::utf_decoder<pugi::impl::utf
16_writer, pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, uns
igned int, unsigned short*)+162>
   0x00407fa2 <+66>:    lea    0x2(%edx),%ebx
   0x00407fa5 <+69>:    cmp    %ebx,%esi
   0x00407fa7 <+71>:    jbe    0x408010 <pugi::impl::utf_decoder<pugi::impl::utf
16_writer, pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, uns
igned int, unsigned short*)+176>
   0x00407fa9 <+73>:    movzwl 0x2(%edx),%edi
   0x00407fad <+77>:    movzwl %di,%ebp
   0x00407fb0 <+80>:    sub    $0xdc00,%ebp
   0x00407fb6 <+86>:    cmp    $0x3ff,%ebp
   0x00407fbc <+92>:    ja     0x408014 <pugi::impl::utf_decoder<pugi::impl::utf
16_writer, pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, uns
igned int, unsigned short*)+180>
   0x00407fbe <+94>:    and    $0x3ff,%ecx
   0x00407fc4 <+100>:   and    $0x3ff,%edi
   0x00407fca <+106>:   add    $0x4,%edx
   0x00407fcd <+109>:   sub    $0x2800,%cx
   0x00407fd2 <+114>:   sub    $0x2400,%di
   0x00407fd7 <+119>:   add    $0x4,%eax
   0x00407fda <+122>:   mov    %cx,-0x4(%eax)
   0x00407fde <+126>:   mov    %di,-0x2(%eax)
   0x00407fe2 <+130>:   cmp    %edx,%esi
   0x00407fe4 <+132>:   ja     0x407f80 <pugi::impl::utf_decoder<pugi::impl::utf
16_writer, pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, uns
igned int, unsigned short*)+32>
   0x00407fe6 <+134>:   pop    %ebx
   0x00407fe7 <+135>:   pop    %esi
   0x00407fe8 <+136>:   pop    %edi
   0x00407fe9 <+137>:   pop    %ebp
   0x00407fea <+138>:   ret
   0x00407feb <+139>:   nop
   0x00407fec <+140>:   lea    0x0(%esi,%eiz,1),%esi
   0x00407ff0 <+144>:   mov    %cx,(%eax)
   0x00407ff3 <+147>:   add    $0x2,%edx
   0x00407ff6 <+150>:   add    $0x2,%eax
   0x00407ff9 <+153>:   cmp    %edx,%esi
   0x00407ffb <+155>:   ja     0x407f80 <pugi::impl::utf_decoder<pugi::impl::utf
16_writer, pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, uns
igned int, unsigned short*)+32>
   0x00407ffd <+157>:   pop    %ebx
   0x00407ffe <+158>:   pop    %esi
   0x00407fff <+159>:   pop    %edi
   0x00408000 <+160>:   pop    %ebp
   0x00408001 <+161>:   ret
   0x00408002 <+162>:   add    $0x2,%edx
   0x00408005 <+165>:   jmp    0x407ff9 <pugi::impl::utf_decoder<pugi::impl::utf
16_writer, pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, uns
igned int, unsigned short*)+153>
   0x00408007 <+167>:   mov    %esi,%esi
   0x00408009 <+169>:   lea    0x0(%edi,%eiz,1),%edi
   0x00408010 <+176>:   mov    %ebx,%edx
   0x00408012 <+178>:   jmp    0x407ff9 <pugi::impl::utf_decoder<pugi::impl::utf
16_writer, pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, uns
igned int, unsigned short*)+153>
   0x00408014 <+180>:   mov    %edi,%ecx
   0x00408016 <+182>:   mov    %ebx,%edx
   0x00408018 <+184>:   jmp    0x407f83 <pugi::impl::utf_decoder<pugi::impl::utf
16_writer, pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, uns
igned int, unsigned short*)+35>
End of assembler dump.
(gdb)

Original comment by raptorfactor@raptorfactor.com on 3 Dec 2013 at 11:04

[GoogleCodeExporter]

I've dumped the disassembly for both the original and the 'fixed' variants of 
the function. Sorry, I'm not sure which syntax you prefer so I've used the 
Intel syntax this time.

Haven't had a look at it yet, but at least it should be easy to compare now.

Original:

========== START ORIGINAL DUMP ==========
(gdb)disassemble 0x407f60
Dump of assembler code for function 
pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) :
0x00407f60 <+0> : push   ebp
0x00407f61 <+1> : push   edi
0x00407f62 <+2> : push   esi
0x00407f63 <+3> : push   ebx
0x00407f64 <+4> : mov    edx, DWORD PTR[esp + 0x14]
0x00407f68 <+8> : mov    ecx, DWORD PTR[esp + 0x18]
0x00407f6c <+12> : mov    eax, DWORD PTR[esp + 0x1c]
0x00407f70 <+16> : lea    esi, [edx + ecx * 2]
0x00407f73 <+19> : cmp    edx, esi
0x00407f75 <+21> : jae    0x407fe6 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 134>
0x00407f77 <+23> : mov    esi, esi
0x00407f79 <+25> : lea    edi, [edi + eiz * 1 + 0x0]
0x00407f80 <+32> : movzx  ecx, WORD PTR[edx]
0x00407f83 <+35> : cmp    cx, 0xd7ff
0x00407f88 <+40> : jbe    0x407ff0 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 144>
0x00407f8a <+42> : movzx  ebx, cx
0x00407f8d <+45> : sub    ebx, 0xe000
0x00407f93 <+51> : cmp    ebx, 0x1fff
0x00407f99 <+57> : jbe    0x407ff0 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 144>
0x00407f9b <+59> : cmp    cx, 0xdbff
0x00407fa0 <+64> : ja     0x408002 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 162>
0x00407fa2 <+66> : lea    ebx, [edx + 0x2]
0x00407fa5 <+69> : cmp    esi, ebx
0x00407fa7 <+71> : jbe    0x408010 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 176>
0x00407fa9 <+73> : movzx  edi, WORD PTR[edx + 0x2]
0x00407fad <+77> : movzx  ebp, di
0x00407fb0 <+80> : sub    ebp, 0xdc00
0x00407fb6 <+86> : cmp    ebp, 0x3ff
0x00407fbc <+92> : ja     0x408014 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 180>
0x00407fbe <+94> : and    ecx, 0x3ff
0x00407fc4 <+100> : and    edi, 0x3ff
0x00407fca <+106> : add    edx, 0x4
0x00407fcd <+109> : sub    cx, 0x2800
0x00407fd2 <+114> : sub    di, 0x2400
0x00407fd7 <+119> : add    eax, 0x4
0x00407fda <+122> : mov    WORD PTR[eax - 0x4], cx
0x00407fde <+126> : mov    WORD PTR[eax - 0x2], di
0x00407fe2 <+130> : cmp    esi, edx
0x00407fe4 <+132> : ja     0x407f80 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 32>
0x00407fe6 <+134> : pop    ebx
0x00407fe7 <+135> : pop    esi
0x00407fe8 <+136> : pop    edi
0x00407fe9 <+137> : pop    ebp
0x00407fea <+138> : ret
0x00407feb <+139> : nop
0x00407fec <+140> : lea    esi, [esi + eiz * 1 + 0x0]
0x00407ff0 <+144> : mov    WORD PTR[eax], cx
0x00407ff3 <+147> : add    edx, 0x2
0x00407ff6 <+150> : add    eax, 0x2
0x00407ff9 <+153> : cmp    esi, edx
0x00407ffb <+155> : ja     0x407f80 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 32>
0x00407ffd <+157> : pop    ebx
0x00407ffe <+158> : pop    esi
0x00407fff <+159> : pop    edi
0x00408000 <+160> : pop    ebp
0x00408001 <+161> : ret
0x00408002 <+162> : add    edx, 0x2
0x00408005 <+165> : jmp    0x407ff9 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 153>
0x00408007 <+167> : mov    esi, esi
0x00408009 <+169> : lea    edi, [edi + eiz * 1 + 0x0]
0x00408010 <+176> : mov    edx, ebx
0x00408012 <+178> : jmp    0x407ff9 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 153>
0x00408014 <+180> : mov    ecx, edi
0x00408016 <+182> : mov    edx, ebx
0x00408018 <+184> : jmp    0x407f83 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 35>
End of assembler dump.
(gdb)
========== STOP ORIGINAL DUMP ==========

========== START FIXED DUMP ==========
(gdb)disas 0x407f60
Dump of assembler code for function 
pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsignedint, 
unsigned short*) :
0x00407f60 <+0> : push   ebp
0x00407f61 <+1> : push   edi
0x00407f62 <+2> : push   esi
0x00407f63 <+3> : push   ebx
0x00407f64 <+4> : mov    edx, DWORD PTR[esp + 0x14]
0x00407f68 <+8> : mov    ecx, DWORD PTR[esp + 0x18]
0x00407f6c <+12> : mov    eax, DWORD PTR[esp + 0x1c]
0x00407f70 <+16> : lea    esi, [edx + ecx * 2]
0x00407f73 <+19> : cmp    edx, esi
0x00407f75 <+21> : jae    0x407fff 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 159>
0x00407f7b <+27> : nop
0x00407f7c <+28> : lea    esi, [esi + eiz * 1 + 0x0]
0x00407f80 <+32> : movzx  ebx, WORD PTR[edx]
0x00407f83 <+35> : movzx  ecx, bx
0x00407f86 <+38> : cmp    ecx, 0xd7ff
0x00407f8c <+44> : jbe    0x408004 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 164>
0x00407f8e <+46> : lea    edi, [ecx - 0xe000]
0x00407f94 <+52> : cmp    edi, 0x1fff
0x00407f9a <+58> : jbe    0x408004 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 164>
0x00407f9c <+60> : sub    ecx, 0xd800
0x00407fa2 <+66> : cmp    ecx, 0x3ff
0x00407fa8 <+72> : ja     0x408020 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 192>
0x00407faa <+74> : lea    ecx, [edx + 0x2]
0x00407fad <+77> : cmp    esi, ecx
0x00407faf <+79> : jbe    0x408025 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 197>
0x00407fb1 <+81> : movzx  edi, WORD PTR[edx + 0x2]
0x00407fb5 <+85> : movzx  ebp, di
0x00407fb8 <+88> : sub    ebp, 0xdc00
0x00407fbe <+94> : cmp    ebp, 0x3ff
0x00407fc4 <+100> : ja     0x408030 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 208>
0x00407fc6 <+102> : and    ebx, 0x3ff
0x00407fcc <+108> : and    edi, 0x3ff
0x00407fd2 <+114> : add    edx, 0x4
0x00407fd5 <+117> : shl    ebx, 0xa
0x00407fd8 <+120> : add    eax, 0x4
0x00407fdb <+123> : lea    ecx, [edi + ebx * 1]
0x00407fde <+126> : mov    ebx, ecx
0x00407fe0 <+128> : and    ecx, 0x3ff
0x00407fe6 <+134> : shr    ebx, 0xa
0x00407fe9 <+137> : sub    cx, 0x2400
0x00407fee <+142> : sub    bx, 0x2800
0x00407ff3 <+147> : mov    WORD PTR[eax - 0x2], cx
0x00407ff7 <+151> : mov    WORD PTR[eax - 0x4], bx
0x00407ffb <+155> : cmp    esi, edx
0x00407ffd <+157> : ja     0x407f80 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 32>
0x00407fff <+159> : pop    ebx
0x00408000 <+160> : pop    esi
0x00408001 <+161> : pop    edi
0x00408002 <+162> : pop    ebp
0x00408003 <+163> : ret
0x00408004 <+164> : mov    WORD PTR[eax], bx
0x00408007 <+167> : add    edx, 0x2
0x0040800a <+170> : add    eax, 0x2
0x0040800d <+173> : cmp    esi, edx
0x0040800f <+175> : ja     0x407f80 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 32>
0x00408015 <+181> : pop    ebx
0x00408016 <+182> : pop    esi
0x00408017 <+183> : pop    edi
0x00408018 <+184> : pop    ebp
0x00408019 <+185> : ret
0x0040801a <+186> : lea    esi, [esi + 0x0]
0x00408020 <+192> : add    edx, 0x2
0x00408023 <+195> : jmp    0x40800d 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 173>
0x00408025 <+197> : mov    edx, ecx
0x00408027 <+199> : jmp    0x40800d 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 173>
0x00408029 <+201> : lea    esi, [esi + eiz * 1 + 0x0]
0x00408030 <+208> : mov    ebx, edi
0x00408032 <+210> : mov    edx, ecx
0x00408034 <+212> : jmp    0x407f83 
<pugi::impl::utf_decoder<pugi::impl::utf16_writer, 
pugi::impl::opt_false>::decode_utf16_block(unsigned short const*, unsigned int, 
unsigned short*) + 35>
End of assembler dump.
(gdb)
========== STOP FIXED DUMP ==========

Original comment by raptorfactor@raptorfactor.com on 4 Dec 2013 at 3:07

[GoogleCodeExporter]

Disassembly looks the same (modulo instruction reordering/extra nop-s) and 
correct, so I guess it's just a badly worded warning.

It's weird though that the warning only appears in one function despite the 
fact that a number of places use the same pattern. I'll try to build gcc 4.9 to 
test this myself just to make sure that all configurations build cleanly with 
-Wstrict-overflow.

Original comment by arseny.k...@gmail.com on 4 Dec 2013 at 3:28

[GoogleCodeExporter]

Finally built gcc 4.9 and verified the change. Fixed in r957.

Original comment by arseny.k...@gmail.com on 20 Dec 2013 at 8:27

• Changed state: Fixed

[GoogleCodeExporter]

Thanks!

Original comment by raptorfactor@raptorfactor.com on 20 Dec 2013 at 9:05