Here are some ideas:
1) Add rules to your makefile to postprocess the tool output and remove the
timestamps before CVS sees them.
2) I've never tried this, but perhaps cvswrappers can remove the timestamps
for you. Even if this works, I bet "cvs -n up" would still report lots of
spurious modified files.
3) Here's a particularly ugly hack; but desperate times call for desperate
measures. Perhaps you can rig up your tool output so that the timestamp
appears to be within a $Log$ keyword expansion. I don't know, but perhaps
"cvs -n up" ignores differences due solely to keyword expansions.
However, I had the impression that "cvs -n up" identifies files as being
modified based on a discrepancy in the file's actual modification date and
what is stored in CVS/Entries; not on whether or not the file matches
what's in the repository. If this is true, then the timestamp doesn't
matter, simply rebuilding your C++ files will make CVS think they've
been modified.
Jonathan M. Gilligan wrote:
>Here's the rub---the tool puts a time-stamp in the C++ files when it
>generates them, so I end up with lots of files that differ from the files
>in the repository only by one line---the one with the time stamp. This
>makes it hard to use "cvs -n up" to track which files are out of date with
>respect to the repository and which have merely had their time-stamps
updated.
