Re: [lldb-dev] Optional Dependencies in LLDB
On Mon, 6 Jan 2020, Jonas Devlieghere via lldb-dev wrote: I just wanted to let you know that most of the work is complete for auto-detecting optional dependencies in LLDB. Unless explicitly specified, optional dependencies like editline will be enabled when available and disabled otherwise. This is different from the old behavior, where optional dependencies were that were enabled by default would cause an error at configuration time. The motivation is to make it easier to build LLDB by making things "just work" out of the box. I think (didn't test at the moment, just browsed the cmakefiles) one case that still isn't handled properly, is the interaction between python/lua and swig. If python or lua are detected, they are enabled, and then the build strictly requires SWIG to be present. I think we should check for SWIG first and require it to be present before automatically enabling python and lua. // Martin ___ lldb-dev mailing list lldb-dev@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-dev
Re: [lldb-dev] Optional Dependencies in LLDB
On Tue, Jan 7, 2020 at 12:52 PM Martin Storsjö wrote: > > On Mon, 6 Jan 2020, Jonas Devlieghere via lldb-dev wrote: > > > I just wanted to let you know that most of the work is complete for > > auto-detecting optional dependencies in LLDB. Unless explicitly > > specified, optional dependencies like editline will be enabled when > > available and disabled otherwise. This is different from the old > > behavior, where optional dependencies were that were enabled by > > default would cause an error at configuration time. The motivation is > > to make it easier to build LLDB by making things "just work" out of > > the box. > > I think (didn't test at the moment, just browsed the cmakefiles) one case > that still isn't handled properly, is the interaction between python/lua > and swig. If python or lua are detected, they are enabled, and then the > build strictly requires SWIG to be present. Yup, good point, that's still on my TODO list. > I think we should check for SWIG first and require it to be present before > automatically enabling python and lua. That would make sense, but I haven't gone that route because downstream (Swift) we have the Python bindings checked-in. So it's not necessary to have SWIG in order to enable Python. Of course downstream shouldn't direct what we do upstream, but if I can figure out a solution that minimizes divergence I strongly prefer that. I'm hoping to get to this soon. > > // Martin > ___ lldb-dev mailing list lldb-dev@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-dev
Re: [lldb-dev] Optional Dependencies in LLDB
After trying it out I concluded that it should be easy enough to check for the static bindings flag in FindPythonInterpAndLibs.cmake so I've implemented your suggestion in fc6f15d4d2c. Thanks again for bringing this up. On Tue, Jan 7, 2020 at 1:01 PM Jonas Devlieghere wrote: > > On Tue, Jan 7, 2020 at 12:52 PM Martin Storsjö wrote: > > > > On Mon, 6 Jan 2020, Jonas Devlieghere via lldb-dev wrote: > > > > > I just wanted to let you know that most of the work is complete for > > > auto-detecting optional dependencies in LLDB. Unless explicitly > > > specified, optional dependencies like editline will be enabled when > > > available and disabled otherwise. This is different from the old > > > behavior, where optional dependencies were that were enabled by > > > default would cause an error at configuration time. The motivation is > > > to make it easier to build LLDB by making things "just work" out of > > > the box. > > > > I think (didn't test at the moment, just browsed the cmakefiles) one case > > that still isn't handled properly, is the interaction between python/lua > > and swig. If python or lua are detected, they are enabled, and then the > > build strictly requires SWIG to be present. > > Yup, good point, that's still on my TODO list. > > > I think we should check for SWIG first and require it to be present before > > automatically enabling python and lua. > > That would make sense, but I haven't gone that route because > downstream (Swift) we have the Python bindings checked-in. So it's not > necessary to have SWIG in order to enable Python. Of course downstream > shouldn't direct what we do upstream, but if I can figure out a > solution that minimizes divergence I strongly prefer that. I'm hoping > to get to this soon. > > > > > // Martin > > ___ lldb-dev mailing list lldb-dev@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-dev
Re: [lldb-dev] Optional Dependencies in LLDB
On Tue, 7 Jan 2020, Jonas Devlieghere wrote: After trying it out I concluded that it should be easy enough to check for the static bindings flag in FindPythonInterpAndLibs.cmake so I've implemented your suggestion in fc6f15d4d2c. Thanks again for bringing this up. Awesome, thanks! Do I understand this correctly that something similar still would be needed for Lua though? // Martin ___ lldb-dev mailing list lldb-dev@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-dev
