diff options
author | Boris Kolpackov <boris@codesynthesis.com> | 2024-02-22 07:12:02 +0200 |
---|---|---|
committer | Boris Kolpackov <boris@codesynthesis.com> | 2024-02-22 07:12:02 +0200 |
commit | 73b82917b76243ab0c8ede3cd12bdbe4fe5ffb08 (patch) | |
tree | ea3467b4828f43baf1d8d0890a34ee95006925c5 /doc | |
parent | bb3e24a35765ab277a8d4fad4bca40f6c0dc6754 (diff) |
Further work on packaging guide
Diffstat (limited to 'doc')
-rw-r--r-- | doc/packaging.cli | 124 |
1 files changed, 113 insertions, 11 deletions
diff --git a/doc/packaging.cli b/doc/packaging.cli index aef8220..fba421d 100644 --- a/doc/packaging.cli +++ b/doc/packaging.cli @@ -21,6 +21,9 @@ // // @@ Note on library metadata where talk about configuration. Also about // autoconf. +// +// @@ Add link from metadata HOWTO to "Modifying upstream source code with +// preprocessor". " \h0#preface|Preface| @@ -703,7 +706,7 @@ public headers are included and where they are installed. The two common \i{good} practices is to either include the public headers with a library name as a subdirectory, for example, \c{#include\ <foo/util.h>}, or to include the library name into each public header name, for example, \c{#include\ -<foo_util.h>} or \c{#include\ <foo.h>} (in the last example the header name is +<foo-util.h>} or \c{#include\ <foo.h>} (in the last example the header name is the library name itself, which is also fairly common). Unfortunately, there is also a fairly common \i{bad} practice: having generically named headers (such as \c{util.h}) included without the library name as a subdirectory. @@ -729,13 +732,13 @@ with the above example, a header that is included as \c{<foo/util.h>} would normally be installed as \c{/usr/include/foo/util.h}. On the other hand, if the library name is part of the header name, then the headers are usually (but not always) installed directly into, say, \c{/usr/include/}, for example as -\c{/usr/include/foo_util.h}. +\c{/usr/include/foo-util.h}. \N|While these are the commonly used installation schemes, there are deviations. In particular, in both cases upstream may choose to add an -additional subdirectory when installing (so the above examples we instead end -up with, say, \c{/usr/include/foo_v1/foo/util.h} and -\c{/usr/include/foo_v1/sub/foo_util.h}). See +additional subdirectory when installing (so the above examples will instead +end up with, say, \c{/usr/include/foo-v1/foo/util.h} and +\c{/usr/include/foo-v1/sub/foo-util.h}). See \l{#howto-extra-header-install-subdir How do I handle extra header installation subdirectory} if you encounter such a case.| @@ -1745,7 +1748,7 @@ Normally the only change that you would make to this \c{buildfile} is to adjust the installation location of headers (see \l{b#intro-operations-install Installing} for background). In particular, if our headers were included without the \c{<foo/...>} prefix but instead contained the library name in -their names (for example, \c{foo_util.hpp}), then the installation setup would +their names (for example, \c{foo-util.hpp}), then the installation setup would instead look like this: \ @@ -3759,6 +3762,11 @@ Also don't forget about tests, examples, etc., which may also add new or remove old source files (typically new tests). See \l{#core-test-upstream-convert Convert smoke test to upstream tests}. +If there are any manual modifications to the upstream source code, then you +will also need to re-apply them to the new version as discussion in +\l{#howto-patch-upstream-source-manual Modifying upstream source code +manually}. + \h2#core-version-management-new-version-build|New version: changes to build system| @@ -4267,8 +4275,6 @@ $ diff -u foo.cpp.orig foo.cpp >foo.cpp.patch || -@@ Add entry/link from New version steps. - \h2#howto-patch-upstream-source-build|Modifying upstream source code during build| @@ -4308,8 +4314,60 @@ in{config.h.in}: file{config.h.cmake} \h2#howto-patch-upstream-source-preproc|Modifying upstream source code with preprocessor| -@@@ +A good illustration of this approach is adding the \c{build2} metadata to an +executable (see +\l{https://github.com/build2/HOWTO/blob/master/entries/convey-additional-information-with-exe-lib.md +How do I convey additional information (metadata) with executables and C/C++ +libraries?} for background). Let's say we have a symlink to upstream's +\c{main.c} that implements the executable's \c{main()} function and we need to +add a snipped of code at the beginning of this function that handles the +\c{--build2-metadata} option. While manually modifying \c{main.c} is not a +wrong approach, we can try to be clever and do it automatically with a +preprocessor. +Specifically, we can create another file next to the \c{main.c} symlink, +calling it, for example, \c{main-build2.c}, with the following content: + +\ +/* Handle --build2-metadata in main() (see also buildfile). */ + +#define main xmain +#include \"main.c\" +#undef main + +#include <stdio.h> +#include <string.h> + +int main (int argc, const char** argv) +{ + if (argc == 2 && strncmp (argv[1], \"--build2-metadata=\", 18) == 0) + { + printf (\"# build2 buildfile foo\n\"); + printf (\"export.metadata = 1 foo\n\"); + printf (\"foo.name = [string] foo\n\"); + ... + return 0; + } + + return xmain (argc, argv); +} +\ + +The idea here is to rename the original \c{main()} with the help of the C +preprocessor and provide our own \c{main()} which, after handling +\c{--build2-metadata} calls the original. One notable deal-breaker for this +approach are C++ implementations of \c{main()} that don't have an explicit +\c{return}. There is also a better chance in C++ for the \c{main} macro to +replace something unintended. + +To complete this we also need to exclude \c{main.c} from compilation in +our \c{buildfile} (since it is compiled as part of \c{main-build2.c} via +the preprocessor inclusion). For example: + +\ +exe{foo}: {h c}{** -main} +exe{foo}: c{main}: include = adhoc # Included in main-build2.c. +\ \h#howto-bad-inclusion-practice|How do I deal with bad header inclusion practice| @@ -4320,14 +4378,58 @@ libraries cannot coexist, neither in the same build nor when installed. For background and details, see \l{intro#proj-struct Canonical Project Structure}. @@ TODO +@@ Dual? \h#howto-extra-header-install-subdir|How do I handle extra header installation subdirectory| This sections explains how to handle an additional header installation -subdirectory. +subdirectory. As an illustration of the problem, consider the \c{libfoo} +example from the previous sections (see the \l{#core-fill-source Fill with +upstream source code} step for a refresher). In that example the library +headers are included as \c{<foo/util.hpp>} and installed as, say, +\c{/usr/include/foo/util.hpp}. In this scheme the installed header inclusion +works without requiring any extra steps from our side because the compiler +searches for header in \c{/usr/include} by default. + +However, some libraries choose to install their headers into a subdirectory +of, say, \c{/usr/include} but without having this subdirectory as part of the +inclusion path (like \c{foo/} in \c{<foo/util.hpp>}). The two typical reasons +for this are support for installing multiple versions of the same library +side-by-side (for example, \c{/usr/include/foo-v1/foo/util.hpp}) and not using +the library name as the inclusion directory prefix and then having to hide the +headers in a subdirectory due to potential clashes with other headers (if +installed directly into, say, /c{/usr/include}). + +In such cases the installed header inclusion does not work out of the box and +we have to arrange for an additional header search path to be added via +\c{pkg-config}. Let's use the versioned library case to illustrate this +technique. The relevant part from the \l{#core-adjust-build-src-header header +\c{buildfile}} will look like this: + +\ +# Install into the foo-vN/foo/ subdirectory of, say, /usr/include/ +# recreating subdirectories. +# +{hxx ixx txx}{*}: +{ + install = include/\"foo-v$version.major\"/foo/ + install.subdirs = true +} +\ -@@ TODO +The part that we need to add, this time to the +\l{#core-adjust-build-src-source source \c{buildfile}}, looks like this: + +\ +# Make sure headers installed into, say, /usr/include/foo-vN/foo/ +# can be included as <foo/*.hpp> by overriding the header search +# path in the generated pkg-config files. +# +lib{foo}: cxx.pkgconfig.include = include/\"foo-v$version.major\"/ +\ + +\N|The variable will be \c{c.pkgconfig.include} for a C library.| \h#howto-no-extension-header|How do I handle headers without extensions| |