Reid Spencer wrote:> It looks to me like your libLLVMSupport.a didn't get built or you didn't > update all of LLVM. Please make sure you update from the root of the > tree with the command: > > cvs update -ARPd > > Reid. >Thanks for the quick response. I have resolved my problem already - basically I had "-s" in my CFLAGS/CXXFLAGS environment which configure put in the Makefiles and screwed up the build. libLLVMSupport did get build, with the IncludeFile.o object (verified with nm), but with -s it confused ld -r at some point! (Perhaps a configure check can be added for this flag to save future hair-pulling) In any case, I removed the CFLAGS variable and am currently building (it seems to be going well). Thanks for the help, LL
On Sat, 2006-06-24 at 18:46 +0100, Lexington Luthor wrote:> I have resolved my problem already - basically I had "-s" in my > CFLAGS/CXXFLAGS environment which configure put in the Makefiles and > screwed up the build.You mean the Makefile system honored your request to use non-standard CFLAGS and CXXFLAGS. The build went fine. It just didn't produce what you were hoping for.> libLLVMSupport did get build, with the > IncludeFile.o object (verified with nm), but with -s it confused ld -r > at some point!The -s option tells the linker to strip *all* symbol information from the file, not just debug symbols. Use -S (capital S) to remove debug symbols. When you strip all symbols, there's nothing to link against and I would expect it to produce the result you got, or worse.> (Perhaps a configure check can be added for this flag to > save future hair-pulling)There are hundreds of options that can screw up a build. When developer's stray from the recommended build procedures, they will have to deal with the consequences themselves. I'm sorry for the trouble it caused, but there isn't much we can do to prevent you from shooting yourself in the foot. :) I've been there, done that kind of thing before and know about the wasted time, etc. I usually just chalk it up to a learning experience. In any event, what were you trying to accomplish with the use of the -s flag? Perhaps there's another way to do it.> > In any case, I removed the CFLAGS variable and am currently building (it > seems to be going well).Great! Glad its working now. Reid. -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 189 bytes Desc: This is a digitally signed message part URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20060624/ee7b27f6/attachment.sig>
Reid Spencer wrote:> I'm sorry for the trouble it caused, but there isn't much we can do to > prevent you from shooting yourself in the foot. :) I've been there, > done that kind of thing before and know about the wasted time, etc. I > usually just chalk it up to a learning experience. > > In any event, what were you trying to accomplish with the use of the -s > flag? Perhaps there's another way to do it. >Heh, no I didn't add it there on purpose. The whole reason I am playing with LLVM is to try coax every last bit of performance out of this program I have been using a lot lately, and among my attempts to do so, I played with gcc flags, which I forgot to unset from my environment when I went on to build LLVM... silly mistake on my part. (I am not in the habit of running stripped binaries, its just that for a program that takes 26 minutes to run each time, even a 1% speed boost is very welcome, which is why I employed more aggressive compiler optimizations.) Anyway, LLVM is fantastic. Thanks to all the developers for the great work, and sorry for the noise on the mailing list. My problem has been resolved and I learned a lot about LLVM in the process. Now onto cachegrind... Thanks for the help, LL