On Mon, Sep 26, 2016 at 3:36 AM, Daniel Shahaf <d...@daniel.shahaf.name> wrote: > Lorenz wrote on Mon, Sep 26, 2016 at 06:01:31 +0000: >> are you sure about tha being a bug? >> >> If for instance in the file the URL keyword is used to initialize a >> string variable, wouldn't you want the file to be recompiled after the >> switch? > > You are describing a different scenario than the OP. > > In the OP's scenario, the file content with keywords expanded was the > same before and after the switch, yet the mtime differed. I consider > that a bug.
How can Subversion know that? Reprocessing the file, setting aside the processed file, then doing a "diff" between them is a lot more pain than it's worth. And the maintenance of supporting such a feature for all releases, and backporting it to previous Subversion releases to support such a contorted workflow as switching upstream midrepos or branches midstream, is potentially quite fragile. If you switch the upstream repo for a software source repo, recompile. It's the best way to make sure you've not left some datestamped file inconsistencies that could mess with the system. > If the file content with keywords expanded had been different before the > switch to after to the switch, then yes, I would have expected the mtime > to differ, too. And how, exactly, are you going to verify this? Put in a locally processed working file to check against? That's begging to leave debris in the working repository./ > Thanks for contributing this observation: the key question is whether > the translated content was modified, not whether the repository-normal > content was. > > Cheers > > Daniel
