our units use dual images for upgrade safety which only differ in address locations. each image uses the same source with project files (.uvproj) that only change the scatter file used for linking. When doing a full compile of each image I've recently started noticing that the images differ by 4 bytes. This appears to happen during linking and compiling on one of our other systems doesn't produce this issue so I suspect something has been corrupted on my laptop. This is very disconcerting. Any ideas??
Could it be that one of the scatter files has "bad" alignment, so it gets adjusted ... ?
the only difference is the base address used in NOR flash. They are on 1 MB boundaries so no. I've also tried using the first image's scatter for the second image with no difference. The second always comes out 4 bytes larger. I'm not really sure which is actually at fault. Is the first clipped for some reason or does the second have extra rubbish. The one thing outside of the keil tools is that I use cygwin's objcopy to create the final binaries, so something could be at fault with that. Just haven't had the time to dig into it yet and though maybe someone would have have heard of this before. A few years back we were having a more major size difference issue, but that turned out to be one of the keil options (I forget which it was) that would perform mutiple compiles on the same build to optimize something, but that's not the problem here.
Thanks
Compare the map files to see where, exactly, the difference happens ... ?
That would be the first logical step in determining the fault. It should ideally be done before even considering a post about it on a forum.
Then, are you using the --feedback option? if so, you must really do multiple rebuilds (typically 2 or 3 is sufficient) before it fully determines what can be completely removed.
To trolls like you that have nothing better to do than watch a forum for people to write smart ass useless replies to, try reading what you are responding to and see that your stupid remarks were meaningless because I had already used the same scatter for both builds and still had the issue and that I wasn't using the "feedback" option. I'm not one of your newbie punks that you can intimidate with your B.S. I've been doing this for over 20 years twit. Idiots like you should be banned from forums.
1) Did I mention scatter? NO 2) Did you previously mention not using --feedback? NO
Seems odd that someone having 20 years experience feels it necessary to post to a forum like this without properly investigating the basics and not providing all useful information. No?
Literals might pool or be required differently based on address. ie bit patterns allow or disallow immediates
again. tried using the same scatter file for both (i.e. same addressing) and the second project file is a copy of the first with just pre & post processing files switched. Pre, normally just copies over the the correct scatter to a file with a common name and post is a lot of moving files about and format conversions which is why I wondered about the cygwin objcopy which is just converting a hex file to a binary and like I've said running on our official build machine produces the correct result. Thanks at least for giving a thought out answer unlike the previous trolls smart ass responses. I should probably try a full reinstall of uvision since it seems more like an environment corruption issue. Just thought maybe someone had run into this before and could maybe get a quick answer. This product is 7 years old and other than the one time because of an option chosen this is the first time this has happened.
Comparing the two configurations, where exactly do the files deviate in size? Is it at the end of the Keil tools or after the cygwin stuff? What have the MAP files told you? Have you produced compiler listing files for the two builds and compared them?
Been there, done that with nothing conclusive, only the general area it's in. Just haven't had enough time for digging any farther especially since it doesn't occur on our official build machine.
Like the last reply, Thanks for thought out response instead of smart-ass troll replies.
Well, a reinstall seems to have cleared up the issue. Something must have been corrupted in the environment. Thanks to those who tried to give worthwhile answers.
Pity you chose to use such a broad brush to narrow down the situation. Nothing of value learned :(