|
From: Dallman, J. <jg...@ug...> - 2006-01-26 14:36:01
|
> Do you guys think that if I used wrapmsvc (some wrapper=20 > for msvc to use gcc, translating arguments from msvc to=20 > gnu type), will valgrind work? No. I'm impressed by the degree to which you've misunderstood=20 the issues involved, though. From a quick Google and your=20 comments, wrapmsvc, is a tool for translating gcc compilation=20 options into the corresponding msvc ones, so that one can use=20 msvc with makefiles set up for gcc. Is that right? If so, this=20 about equivalent to teaching someone to swim by putting them=20 into a swimming costume and claiming you've finished.=20 What will matter to Valgrind is what the compiler produces, rather than how you ask the compiler to do things. And=20 Valgrind only understands programs that are built for Linux (and very similar operating systems, when built for them).=20 Windows is a very different operating system indeed, and Valgrind does not understand Windows issues at all. Yes, it would be useful if it did. But it doesn't, and to make it=20 do so would require a great deal of work. Nobody has come=20 along with both the motivation and the skills to do that as yet.=20 To save your time, here are two other things that won't work: * DJPP, or Cygwin, or other toolsets that allow you to run=20 GCC on Windows for producing Windows programs. This is no=20 help because you'd still be asking Valgrind to work on a=20 Windows program, and it doesn't know how to do that.=20 * WINE, or other Windows emulators for Linux. These allow a Windows program to run on Linux, by surrounding it with a semi-complete simulation of Windows. That still wouldn't=20 let Valgrind deal with a Windows program. You could run=20 Valgrind on the Windows emulator, which would include the Windows program running within it, but the odds of ever learning anything useful from this are very small: it's=20 too complicated.=20 --=20 John Dallman, Parasolid Porting Engineer, +44-1223-371554=20 |
|
From: Dennis L. <pla...@in...> - 2006-01-26 18:41:57
|
At 15:35 26.01.2006, Dallman, John wrote: > > Do you guys think that if I used wrapmsvc (some wrapper > > for msvc to use gcc, translating arguments from msvc to > > gnu type), will valgrind work? > >No. I'm impressed by the degree to which you've misunderstood >the issues involved, though. I guess the misunderstanding here is deliberately. He went to #C++ asking for a memory leak checker and was told that there is valgrind, but there is no way to operate V on windows. 5 Minutes after that the mail arrived on the list. Maybe its a language Problem, so anyone speaking his native language should give it a try. greets Dennis Carpe quod tibi datum est |
|
From: Kevin P. <pu...@pu...> - 2006-01-27 06:13:18
|
Dallman, John wrote: > * WINE, or other Windows emulators for Linux. These allow a > Windows program to run on Linux, by surrounding it with a > semi-complete simulation of Windows. That still wouldn't > let Valgrind deal with a Windows program. You could run > Valgrind on the Windows emulator, which would include the > Windows program running within it, but the odds of ever > learning anything useful from this are very small: it's > too complicated. Actually, there was a point at which this was useful (see http://valgrind.org/downloads/variants.html), though I doubt the fragments are much good to a user at this point. The code was never merged to mainline (I'll assume there was a good reason, as I certainly can't claim to understand what was done to this variant) and given the major architectural changes since 2003 I suspect forward-porting the change would be all but impossible. One would really have to dig through the diff vs. a contemporary valgrind release and figure out what it was actually trying to accomplish, then start over on making that happen. As I have no idea what was really different in it I don't know how hard this would be. But, I'd assume from the fact that it's been 2+ years with nothing but an occasional email list mention of the fact that it once existed (hey, now we've got this year's in) that it that it must be fairly involved. |