Re: [PATCH 10/16] rpcgen: define entrypoint for running new rpcgen impl

On a Wednesday in 2023, Daniel P. Berrangé wrote: > The new program takes the form > > rpcgen [--mode source|header|repr] \ > [--header include] \ > xdr-file output-file > > If '--mode' is not given it parses the XDR file but does not > generate anything, which is useful as a syntax check. The > 'source' mode gives the '.c' file content, while 'header' > gives the '.h' file content. 'repr' gives a representation > of the abstract syntax tree, mostly useful for debugging > the parser. > > If '--header' is given, it is added as a local #include ".." > statement in the output and is valid for either 'header' > or 'source' modes. > > Either 'xdr-file' or 'output-file' can be omitted in which > case they default to stdin/stdout respectively. > > This rpcgen program will directly include the 'config.h' > header in its output. > > Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com&gt; > --- > scripts/rpcgen/main.py | 86 ++++++++++++++++++++++++++++++++++++++++++ > 1 file changed, 86 insertions(+) > create mode 100755 scripts/rpcgen/main.py > > diff --git a/scripts/rpcgen/main.py b/scripts/rpcgen/main.py > new file mode 100755 > index 0000000000..bf4ef38ede > --- /dev/null > +++ b/scripts/rpcgen/main.py > @@ -0,0 +1,86 @@ > +#!/usr/bin/python #!/usr/bin/env python > + > + if args.mode == "header": > + print("/* This file is auto-generated from %s */\n" % args.input, file=outfp) > + print("#include <rpc/rpc.h>", file=outfp) > + for h in args.header: > + print('#include "%s"' % h, file=outfp) > + print("", file=outfp) > + print("#pragma once\n", file=outfp) > + generator = XDRTypeDeclarationGenerator(spec) > + print(generator.visit(), file=outfp) > + generator = XDRMarshallDeclarationGenerator(spec) > + print(generator.visit(), file=outfp) > + elif args.mode == "source": > + print("/* This file is auto-generated from %s */\n" % args.input, file=outfp) > + print("#include <config.h>", file=outfp) > + if args.input.endswith(".x"): > + print('#include "%s.h"' % args.input[:-2], file=outfp) The .x file is in the source dir, while the .h file is in the build dir. Unless I remove the two lines above, I'm getting an error: src/remote/lxc_protocol.c:4:10: fatal error: '../../git/libvirt/src/remote/lxc_protocol.h' file not found #include "../../git/libvirt/src/remote/lxc_protocol.h" ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ And thanks to the --header option, there already is an include of lxc_protocol.h in the file: #include <config.h> #include "../../git/libvirt/src/remote/lxc_protocol.h" #include "lxc_protocol.h" If I use a builddir under libvirt.git, it somehow works: #include <config.h> #include "../src/remote/lxc_protocol.h" #include "lxc_protocol.h"