From: Trevor D. (Twylite) <tw...@cr...> - 2012-12-07 15:13:21
|
Hi, On 2012/12/07 04:45 PM, Donald G Porter wrote: >>> If the nmake crowd can live with fragile warnings not to build/test >>> in paths with spaces, I'm not going to more effort to overrule that. > Ok, I was unaware that this is a documented and accepted limitation > of the nmake builds. Apologies for not having that cultural awareness. To clarify: I can't speak for other developers, but building in paths that have spaces in them is always a pain, and I don't know anyone who does it. The tests should run against an installed Tcl+extensions that is in a path with spaces in it, but I wouldn't expect 'nmake test' to do so. > I am not seriously considering actually disposing of the nmake system. > That was not a plan, that was an expression of how I feel. I think it does warrant serious discussion. The main stumbling point seems to be what to replace it with. If the Windows-based developers could agree on an approach, I think we would have a solution. > All that said, I think the nmake improvements are in a state to put > out RC3 and see what more testing can tell us. Hopefully soon, if > we can get Bug 3588687 to come to a resolution. I've just (this afternoon) done trunk builds of Tcl, Tk, Thread, Itcl, Tdbc, TdbcOdbc, and TdbcSqlite3 (build env is Windows 7 64-bit, building 32-bit binary using MSVC 10). All build and install cleanly, and all packages/extensions load into the installed interp. I haven't run tests yet, but I can confirm that tdbcodbc 'nmake test' doesn't work (it assumes that tdbc.tcl is in the same directory as the tdbc DLL, which isn't a valid assumption in the build environment). I'm not going to try and fix this - there are assumptions layered on assumptions about the locations of different bits. I can run tests/all.tcl through the installed tclsh.exe and get no failure against the Jet database. > To make matters worse, there's a clear expectation that the bundled > packages have nmake builds working too. And that's more than a blind > maintenance task. That's adding something which hasn't been there > before at all. And something which the upstreams have no interest in. Mmm. So as it turns out Sqlite3 (as opposed to tdbc::sqlite3) doesn't have a functional makefile.vc ... which makes building an entire Tcl distribution with Sqlite support using MSVC && !msys rather challenging. Regards, Twylite |