Ralf Wildenhues wrote: > * Bollinger, John C wrote on Mon, Sep 14, 2009 at 05:33:40PM CEST: > > It seems that normally, the generated configure script creates dummy > > *.Po files in the right places, thus masking the absence of Make > rules > > for that purpose. In my case, however, I used Make variables in > > Makefile.am to abbreviate the (long) paths from my project root to my > > source files. > > That's the cause. Automake cannot currently cope with non-literals in > source file names. I've tried before to eliminate this unfortunate > limitation (at least for the 'depfiles' config.status commands); maybe > I should dig that out again sometime. But for now, that's what you'd > need to undo to avoid this bug.
I'm still experimenting with this, but other than the issue with automatic dependency tracking, Automake so far seems to be coping admirably with the non-literals in my source file names. (A nice piece of work, then, if it's exceeding the design goals!) I assume it would have more trouble if I tried to put non-literals in target names, and perhaps my particular use of non-literals is especially benign. I consider myself warned that it is not a supported feature. For the time being, I think I'll stick with my solution of disabling automatic dependency tracking. The readability and maintainability afforded by using Make variables for my directory paths far outweigh the convenience of not having to manually express all my dependencies. Please take this as user feedback, not any kind of criticism. I'm still curious, however, about why Automake places the code for depfile generation into configure / config.status, instead of outputting Make rules for those files. If it's an issue of compatibility with non-GNU Make, then perhaps Automake could do both? It seems to me that then lesser Makes would still behave as they do now, but more enlightened ones could handle additional cases. That wouldn't be a complete solution to my particular issue, but it would be a significant improvement, and it's not really specific to cross-directory builds. Anyway, I think I'm good to go now. Thank you for your time and attention. Best, John -- John C. Bollinger, Ph.D. Computing and X-Ray Scientist Department of Structural Biology St. Jude Children's Research Hospital [email protected] (901) 595-3166 [office] www.stjude.org Email Disclaimer: www.stjude.org/emaildisclaimer
