Hi, several of the patches I submitted last week have not yet been applied, and most of it was really uncontroversial stuff like adding some #includes and some typecasts, fixing alloca for visual C etc. I guess I just wrote too many mails so it was lost in the process, so here is a diff which includes all changes I have made, except those related to hash_map differences since my solution here is just too ugly. I fixed the INT64_MIN as pointed out by Misha Brukman and I made yet another fix for the variable length array in LiveVariables.cpp, this time using alloca. I hope these patches are acceptable, please tell me which ones have to be modified (if any) and I will do it ASAP. Meanwhile I will work on a proper solution to the hash_map incompatibilities... m.
On Mon, 25 Oct 2004, Morten Ofstad wrote:> several of the patches I submitted last week have not yet been applied, > and most of it was really uncontroversial stuff like adding some > #includes and some typecasts, fixing alloca for visual C etc.Oops, sorry about that. :(> I guess I just wrote too many mails so it was lost in the process, so > here is a diff which includes all changes I have made, except those > related to hash_map differences since my solution here is just too ugly.Ok.> I fixed the INT64_MIN as pointed out by Misha Brukman and I made yet > another fix for the variable length array in LiveVariables.cpp, this > time using alloca. I hope these patches are acceptable, please tell me > which ones have to be modified (if any) and I will do it ASAP. Meanwhile > I will work on a proper solution to the hash_map incompatibilities...Sounds great. I applied almost all of the patches: http://mail.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20041025/019806.html http://mail.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20041025/019807.html http://mail.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20041025/019808.html http://mail.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20041025/019809.html http://mail.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20041025/019811.html http://mail.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20041025/019812.html http://mail.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20041025/019813.html http://mail.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20041025/019814.html The two patches I didn't apply are the one that affects the interpreter (I'll let Brian tackle it as he's the interpreter guy), and this one: ==================================================================RCS file: /var/cvs/llvm/llvm/include/llvm/CodeGen/MachineBasicBlock.h,v retrieving revision 1.39 diff -u -r1.39 MachineBasicBlock.h --- include/llvm/CodeGen/MachineBasicBlock.h 1 Sep 2004 22:55:34 -0000 1.39 +++ include/llvm/CodeGen/MachineBasicBlock.h 25 Oct 2004 08:29:24 -0000 @@ -18,6 +18,8 @@ #include "llvm/ADT/GraphTraits.h" #include "llvm/ADT/ilist" #include <iosfwd> +#include <algorithm> +#include <functional> namespace llvm { class MachineFunction; For this change, I'd rather not put those extra #includes into a main header like that. Could you please move the inline methods that need the #includes out-of-line into MachineBasicBlock.cpp, then add the #include's there? If I missed any of the patches or messed them up, please let me know. :) Thanks for the great patches! -Chris -- http://llvm.org/ http://nondot.org/sabre/
Chris Lattner wrote:> The two patches I didn't apply are the one that affects the interpreter > (I'll let Brian tackle it as he's the interpreter guy), and this one: > > ==================================================================> RCS file: /var/cvs/llvm/llvm/include/llvm/CodeGen/MachineBasicBlock.h,v > retrieving revision 1.39 > diff -u -r1.39 MachineBasicBlock.h > --- include/llvm/CodeGen/MachineBasicBlock.h 1 Sep 2004 22:55:34 -0000 1.39 > +++ include/llvm/CodeGen/MachineBasicBlock.h 25 Oct 2004 08:29:24 -0000 > @@ -18,6 +18,8 @@ > #include "llvm/ADT/GraphTraits.h" > #include "llvm/ADT/ilist" > #include <iosfwd> > +#include <algorithm> > +#include <functional> > > namespace llvm { > class MachineFunction; > > For this change, I'd rather not put those extra #includes into a main > header like that. Could you please move the inline methods that need the > #includes out-of-line into MachineBasicBlock.cpp, then add the #include's > there?the <functional> is needed because of this declaration: struct MachineBasicBlock2IndexFunctor : std::unary_function<const MachineBasicBlock*, unsigned> { How can it be moved?? removing the <algorithm> caused a few more files to require including this header, I'm attaching a patch to this message. By the way, I'm getting 2356 warnings when I build the project - most of them are either "type name first seen using 'struct' now seen using 'class'" or "type name first seen using 'class' now seen using 'struct'" -- are there any plans to clean this up? I saw the coding guidelines for LLVM says something about the use of struct vs. class, but it doesn't seem like this is well enforced across the project. My personal opinion is that 'struct' should only be used for something which a plain C compiler would recognize...> If I missed any of the patches or messed them up, please let me know. :) > Thanks for the great patches!Thanks for the support! My diff is almost empty now, so I will start working on integrating LLVM with our project ... m. -------------- next part -------------- An embedded and charset-unspecified text was scrubbed... Name: diff.txt URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20041026/0ff92451/attachment.txt>