Re: hash_map reference and gcc 3.X

[Wei Qin <wqin at EE dot Princeton dot EDU>]

Hi, Ron,

	It is probably the namespace problem. You can check out
http://gcc.gnu.org/onlinedocs/libstdc++/faq/#5_4
for compatibility fix.

Wei

On Tue, 12 Aug 2003, Ron Goodwyn wrote:

> Greetings,
>
> I inherited some code that had last been tested using gcc-2.9.5.3.  Now I
> am attempting to cimpile it using 3.2.2-5 (the revision of the GNU tools
> delivered with RedHat 9.0).  Here is the code snippet that is having
> problems:
>
>
> 32 // hash function on strings
> 33 struct str_hash {
> 34 size_t operator()(const char* str) const {
> 35 size_t hashval = 0;
> 36 for ( ; *str; str++)
> 37 hashval = 5*hashval + *str; // shift and add
> 38
> 39 return hashval;
> 40 }
> 41 };
> 42
> 43
> 44 // equal_to function on strings
> 45 struct str_equal {
> 46 bool operator()(const char* s1, const char* s2) const {
> 47 return (strcmp(s1, s2) == 0);
> 48 }
> 49 };
> 50
> 51
> 52 // convenience template for mapping between name and pointer
> 53 template
> 54 class DRTINameMap : public hash_map
> 56 {
> 57 public:
> 58 DRTINameMap() :
> 59 hash_map() {}
> 60 DRTINameMap(const DRTINameMap& nmap) : hash_map(nmap) {}
> };
>
> Here are the compiler errors:
>
> In file included from DRTI.hh:12,
> from RegionInfo.cc:1:
> DRTI_Factories.hh:54: parse error before `<' token
> DRTI_Factories.hh:59: parse error before `&' token
> DRTI_Factories.hh:59: ISO C++ forbids declaration of `DRTINameMap' with no
> type
> DRTI_Factories.hh: In function `int DRTINameMap(...)':
> DRTI_Factories.hh:59: `int DRTINameMap(...)' redeclared as different kind
> of
> symbol
> DRTI_Factories.hh:54: previous declaration of `template
> class DRTINameMap'
> DRTI_Factories.hh:54: previous non-function declaration `template class DRTINameMap'
> DRTI_Factories.hh:59: conflicts with function declaration `int
> DRTINameMap(...)
> '
> DRTI_Factories.hh:59: parse error before `<' token
> DRTI_Factories.hh:59: only constructors take base initializers
> DRTI_Factories.hh:59: confused by earlier errors, bailing out
> make: *** [RegionInfo.lo] Error 1
>
>
> Does anybody have any idea as to what changed from gcc 2.9 to gcc 3.X that
> could cause this behavior?  Based on my research, similar problems seem to
> arise when you forget to specify std:: however, hash_map is not standard,
> it's an extension.  I'm trying to determine whether the code needs to
> change, or if there is some configuration or compiler problem here.
>
> Many thanks,
> Ron Goodwyn
>
>