| |
| |
| |
|
Page: 1 2
Comments:
<0> hi there
<0> i don't know if this is the right chan to ask..
<0> i'll try :D
<0> i'm getting strange errors compiling a cpp file
<0> linking*
<0> main.cpp:(.text+0x210): undefined reference to `operator new(unsigned int)'
<0> and for delete too
<0> main.cpp:(.text+0x28c): undefined reference to `operator delete(void*)'
<0> and other strange errors i can't understand...
<1> your probably not linking against the standard libraries
<0> what can i try?
<2> use "c++" instead of "gcc"
<3> Hangman[]: either use g++ instead of gcc or add "-lstdc++" to the arguments
<0> yeah i'm using g++
<0> i'll try with -lstdc++
<0> wow it works
<0> :D thanks a lot
<3> strange, using g++ should make it so you don't require -lstdc++
<0> i'm using custom build rules on vc++ i don't know if it's the same thing..
<0> ehm....
<0> undefined reference to `pow'
<0> sin, log, exp...
<3> add -lm
<3> m stands for math
<0> uh ok they disappeared :D tnx
<0> and now...
<0> undefined reference to `operator delete[](void*)'
<0> undefined reference to `operator new[](unsigned int)'
<0> :S
<3> vc++ should have some flag to say to link in c++ mode using -lstdc++ isn't really the cleanest way
<3> try moving the -lstdc++ to the end of the link line
<0> GP2X - 0 error(s), 0 warning(s)
<0> yuhuuuu! thank you
<3> :-)
<0> see ya guys and thank you again ;)
<3> a library always has to appear after any object files or other libraries that use symbols in it
<3> thus why moving -lstdc++ to the end helped
<0> i didn't know the order changed something :|
<0> i'm pretty noob ;P
<3> as long as you're always learning ...
<0> but why the lib has to appear after the files that are using it and not first? :|
<0> before
<0> lol my english ****s :P
<3> I'm not sure exactly why but I think it has to do with how the linker looks for symbols. it probably keeps a list of symbols it needs to find and looks for them as it goes, but never looks for them where it has already been.
<0> this makes sense :)
<4> hi. if porting existing code from borland - is there an compatible AnsiString cl*** one can use directly? i mean, AnsiString is 1-based - is there an existing alike AnsiString-Cl*** for GCC?
<2> 1-based?
<2> c++ does have a standard string cl*** called std::string
<5> is there a doc/readme about the advantages of gcc4.x above gcc3.x ?
<2> Henkie-: first link in the topic
<4> Zordan: yes "abcde" 'a' is at position 1 in Borlands AnsiString-Cl***
<2> oh no there is no such thing ... you could probably wrap std::string with an operator[] that adds 1 though
<5> Zordan, that pages mentions the changes, not entirely what i was looking for
<2> maybe you should elaborate then
<4> Zordan: that means to rewrite all used methods, constructors, ...?
<2> trebor_win: probably makes more sense to substract 1 from all your indices :)
<5> Zordan, ok, in a nutshell, is it (noticeable) faster?
<2> Henkie-: well yes and it says so on the changes page
<5> Zordan, ok, than i apologize, must have missed it when checking out that page
<6> Henkie-: what is the x in 3.x?
<2> a wildcard
<6> i'm aware of that. i ask because i don't notice a difference between 4.x and 3.4, but i do notice a large difference between 3.4 and 3.x for 0 <= x <= 3.
<7> hi guys... i need some help on stdlib
<7> is there a wayto extend or inherit fro basic_string, without completely redefining and reimplementing it?
<8> i wrote some avl functions to complement my existing BST functions, and i want to make them universally compatible with BST
<8> err
<8> damnit
<8> src/hunk.c:331: warning: dereferencing type-punned pointer will break strict-aliasing rules
<8> could i have an elaboration?
<8> i don't know what that means
<9> boris : it means it's probably illegal C
<9> and might result in generated code that does something different from what you expect if you have -fstrict-aliasing on
<9> ( which is turned on by -O2, iirc )
<10> I've noticed a bug in GCC's mudflapth. Here's a simple test app, if anyone wants to take a look: http://rafb.net/paste/results/79bMph87.html
<11> morning
<11> how do I force gcc to use gas and not as (on solaris)
<11> AS and AS_FOR_TARGET work during configure, but afterwards it still uses the sun as
<11> *grom*
<11> --with-as and --with-ld don't do what they should do at all
<12> hello
<13> hi, i'm curious, can anyone tell me the difference between mtune, mcpu and march? it's a little unclear to me from the manpage
<11> should not
<13> be unclear from the manpage?
<11> -mtune=cpu-type
<11> Tune to cpu-type everything applicable about the generated code,
<11> except for the ABI and the set of available instructions.
<11> -march=cpu-type
<11> Generate instructions for the machine type cpu-type. The choices
<11> for cpu-type are the same as for -mtune. Moreover, specifying
<11> -march=cpu-type implies -mtune=cpu-type.
<11> -mcpu=cpu-type
<11> A deprecated synonym for -mtune.
<11> now what's unclear?
<11> mcpu is deprecated for i386
<11> you should use mtune
<13> my manpage doesn't say that about -mcpu
<11> mtune does take care of using everything appropriate of the cpu-type except of using the cpu-type-only instructions
<13> okay
<11> march takes care of using cpu-type-only instructions
<11> might be you have to look into the info page to read that
<11> this is a redhat es
<13> mmm ok, so march=pentium4 would automatically enable sse?
<11> maybe they created a nice gcc for me ;)
<11> +manpage
<11> abraxas: no
<11> abraxas: that's something seperate
<13> oh ok
<11> abraxas: no logic there though ;)
<11> abraxas: I just know that ;)
<13> :) ok
<13> so enabling sse can only be done using -msse ?
<11> you know how to use the info command?
<13> info gcc right? :)
<11> yeah but
<11> it's not that easy to use :/
<11> although one can now use arrows and pgup/pgdown I noticed
<11> graphical clients seem to exist too
<11> anyway
<13> no, this is fine
<11> you might want to look at -mfpmath too
<13> ok
<11> -mfpmatch=sse
<10> Anyone want to look at a bug I found in mudflap: http://rafb.net/paste/results/79bMph87.html
<11> and -msse2
<13> ok, cool
<11> have fun ;)
<13> but all in all btw, it's best to both use -march= and -mtune= ?
<13> and give both an equal value?
<11> hm
<11> read back to what I pasted
<11> for -march
<11> last sentence ;)
<13> ah yes :) my bad
<13> lol
<13> so -march will suffice
<11> yes
<13> that + sse/mmx flags
<13> k, cool
<13> thanx :)
<11> np
<13> why would someone use mtune and not march btw? can you think of any reason?
<11> yes
<11> mtune allows you to tune your code for processor X
<11> but still allow it to be run on a 'previous-generation' processor
<13> aah ok
<13> but march should result in faster code, or at least not slower, right?
<11> no guarantees
<11> it _should_
<11> you better test that though
Return to
#gcc
or
Go to some related
logs:
#css
Adding Packages to an Image bitbake
wingrub ubuntu
Desktop symbolic lynk
#web
#linux
rioutil kubuntu
ircnet The Warlords
#kde
#debian
|
|