On Jun 11, 2014, at 4:01 PM, Reid Kleckner <[email protected]> wrote: > This is a general problem: any subcommand we execute may overflow the system > command line length limit. I'd rather solve it transparently for -cc1 as > well as linker invocations by exposing a command line length estimate from > Support and then estimating our command line length in the driver before > invoking the subcommand, optionally with a response file. I know that Sony > has an internal patch to do this for -cc1 invocations, so it's not just > linker stuff.
Agreed that this is a general problem. Supporting response files for clang internally sounds like a great goal IMO (not to mention, it would *really* help debugging — have you ever had a huge cc1 invocation to pass to lldb? ;-)). Im not certain how you would accommodate the fact that some linkers only support file lists, others support file lists and response files. I have to think about this a bit more, but the fact that not all linkers support response files makes me think that we may have to special case the linker invocation :-(. > https://urldefense.proofpoint.com/v1/url?u=http://reviews.llvm.org/D4104&k=ZVNjlDMF0FElm4dQtryO4A%3D%3D%0A&r=CchYc4lrV44%2BZqxZADw0BQ%3D%3D%0A&m=QJSEm6pZFDZmbT4U5loX1JKIT%2BPausNDEIFHkBEKryE%3D%0A&s=6ee6583a599e3929b9b20fd50ccac0fc7fe5311d66a449349aa1093b1481fa29 > > -- Saleem Abdulrasool abdulras (at) fb (dot) com _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
