I tried compiling using GNUARM toolchain in keil(evaluation). In the output window it shows that it has succesfully created the exe. But when I goto the path where exe is supposed to be placed, I dont find any exe there. what could be the problem?
That's true. But you didn't say "executable", did you? - you specifically said "exe" Anyhow, have you checked for the other possibilities like Christoph suggexted; eg, "bin" or no extension at all? Well, well, well. Someone who reads this forum could observe that communication skills of some contributors are less than ideal for a given context of seeking problem resolution. However, I think berating someone for the use of "exe" versus "executable" will not improve their command of English. Raise above it, will you, please? Why do you notice the splinter in your brother's eye, but do not perceive the wooden beam in your own eye?[Matthew 7:3] Nevertheless, I too have encountered the same problem (in the latest release v3.0, but not in the earlier version of RVMDK): Build target 'Simulator' assembling Startup.s... compiling Serial.c... compiling Blinky.c... linking... ".\Flash\Blinky.elf" - 0 Error(s), 0 Warning(s). and I've searched my drive, but blinky.elf is nowhere. Actualy with listing enabled, there is no listing being generated, either. Is there a possibility that this compiler's message is bogus? If blinky.elf is without an "elf" extention (as you had suggested), what the debugger could do if it expects to be there? One could expect the example supplied with the compiler to be working. But it is not, in this case. BTW, the preview button on message posting generates a blank screen (using Firefox 1.5.0.3) so I am submitting this message "as is".
"Is there a possibility that this compiler's message is bogus?" It is possible that the specific messages you've listed aren't generated by the compiler or linker at all, but by the build tool. In this case, it would be possible that the messages are not quite a true reflection of what's actually happened...
You might find that RVMDK Ver 3.0 does not produce elf format but instead outputs intel if the box is checked. For the abs object file, the output might be Keil's propritary object module with no extension. Can someone correct me. I do not have Ver 3.0 installed on this machine to test. Guess I can try to download the eval on my dial up. So it appears to me, the message is not telling to whole truth. Bradford
"You might find that RVMDK Ver 3.0..." Well, you might - but the question was about GNU, not RV!
The RealView MDK also contains GNU support and we cannot duplicate this problem here. Are you sure that the ELF file is not produced? What GNU version are you using? Can you build other examples using GNU?
I have seen behaviour similar to this with the arm-uclibc GNU tools which use Cygwin. These GNU compiler executables require the cygwin1.dll file to execute to provide the Unix emulation layer. If this is not present (or not locatable) then the build will fail silently. It will appear to have built successfully in the build output window. Try running the GNU compiler yourself from the DOS command prompt as a quick test. If you get an error to the effect that cygwin1.dll is missing then you will have to add the location of where the cygwin1.dll is located to the PATH environment variable.
I am using version 4.1. I compiled 2-3 examples and in none of the cases .elf was generated.
Then please use the GNU version that we have published on this web side. We have tested all examples using this version.
There was an issue with the GCC bundle where it was missing a file. I am not sure this has been corrected yet. The file you need is: C:\ProgramFiles\GNUARM\bin\cygintl3.dll Without it builds silently fail. If you are running the Cygwin environment, the copy in the Cygwin /bin directory can be copied in. Kathy Moser