Bug triage for 2.72 is complete. All bug-autoconf subscribers, thank you for your patience with the flood of low-information email. I intend to make the final release of 2.72 no sooner than Monday, Dec. 18, but no *later* than Friday, Dec. 22. Please consider the 'master' branch to be in release freeze starting now: do not commit any patch to 'master' without first posting it to autoconf-patches and waiting for me to OK it. That said, please don't hesitate to tackle anything in the lists below, even if it's not in the top two categories. Testing is still extremely useful, especially testing on OSes other than Linux, testing with development gcc and/or clang, and testing that goes beyond autoconf's built-in test suite (e.g. pick your favorite project, regenerate its configure script with trunk autoconf, and see if it still probes everything correctly). Here's the breakdown of all open bug reports, in decreasing order of priority: ## Release blockers: 1 bug * #110927: 30 tests fail building on macOS Venture 13.5.1 with Xcode 14.3.1 This is believed to have been fixed, but the fix needs testing. I am testing it right now on the GCC Compile Farm's macOS machine, but it's running an older version of both the operating system (12.6) and of Xcode (14.2). If anyone is able to test with an OS that's a closer match to the bug reporter's machine, please do. (You just need to check out the 'master' branch, bootstrap, and run the testsuite. Note that you will need to install a current version of GNU M4 if you haven't already.) ## Desirable for release: 7 bugs * #110318: autoreconf: support libtoolize being named glibtoolize * #110503: Autoconf 2.70 problem: gkt-doc/gtkdocize is now unconditionally required * #110554: AC_CONFIG_HEADERS doesn't work properly for files with Windows line-endings * #110561: autoconf: store autom4te request keys in sorted order * #110713: Use __STDC__VERSION as first test for C99 compatibility (patch attached) * #110872: m4_warn differs in various ways from its documentation * #110971: busybox mkdir is misdetected These all look easy, and many of them have patches from the reporter. I plan to work through all of them over the next few days. ## Desirable for the release after this one: 13 bugs * #110266 [Important]: Revert regression in Yacc support, and add Bison support * #110286 [Important]: Make it possible to request a specific (non-latest) version of a language standard * #109676 [Normal]: only one "checking whether the AC_LANG compiler works" * #110347 [Normal]: Revise documentation of when configure enters cross-compilation mode. * #110348 [Normal]: Simplify rules for when configure enters cross-compiling mode * #108092 [Minor]: AC_PROG_GREP can select /usr/xpg4/bin/grep under Solaris 9 but it doesn't support long lines * #110416 [Minor]: documentation: ordering of basic macros? * #110612 [Minor]: Re-exec of $as_myself chooses wrong configure script from PATH * #110297 [Wish]: AC_PATH_PROG should accept a basename as an override * #110382 [Wish]: Make confdefs.h idempotent vs compiler warnings * #110394 [Wish]: Support optional use of the C++ compiler in AC_PROG_CXX (and similarly for other languages) * #110431 [Wish]: AC_INIT should accept shell variable expansion in its arguments * #110435 [Wish]: autoreconf: recognize when AM_ICONV is used without the rest of gettext Currently I think all of these should be deferred to 2.73, but ideally no *later* than 2.73. If anyone thinks any of these should *not* be deferred, please say so now. ## "To do eventually" and "blocked": 23 bugs I don't expect any progress in the near future on any of these bugs and I don't think we should commit to any time frame for them. If you see something in these lists that you think should be addressed in the near future, *and* you're prepared to tackle it yourself, please speak up. Whether a bug belongs in one or the other of these categories is unfortunately somewhat fuzzy so I'm going to break them down differently for this message. * #110354: Parallel autotest produces mangled output on Solaris 10 I got dibs on this one. I plan to work on it early next year. But it requires a complete overhaul of the parallel test driver and I don't know how long it's going to take. * #110300: configure >/dev/full does not fail promptly May be easy, or may require an enormous amount of work. * #110746: GNU Autoconf 2.71: F77 name-mangling fails when compiling with mpicxx Should be easy ... for someone who deeply understands Fortran and MPI. * #110478: support for slibtool ans optional replacement for GNU libtool This can't go anywhere until the slibtool project does a bunch more work. After that, the Automake project has to do a bunch of work. We only get involved at the very end. * #108879: Add m4_forbid_pattern for autoconf-archive macros * #110212: Transform pkgdatadir using program-prefix and -suffix * #110395: Makefile generated from Makefile.in unqoute *FLAGS variable twice while ./configure script only once * #110417: Do not generate a back up file when running autoreconf * #110418: Unable to specify build and target perl in cross builds * #110537: AC_SEARCH_LIBS does not honor --with-sysroot= Requesting changes that require coordination with Automake and possibly also the Autoconf Archive, Gnulib, and/or the GNU Coding Standards. Whoever tackles the latter two will also need to understand cross compilation in detail. * #110298: "Substitute and define special characters" test fails on NetBSD * #110307: Upstream: netbsd/patch-lib_autoconf_fortran.m4 * #110308: Upstream: yocto/autoreconf-exclude.patch * #110317: Upstream: oe/autotest-automake-result-format.patch * #110451: autoconf 2.71 generated configure file does not enable C compiler flag for C99 on Solaris 10 * #110497: "This script requires a shell more modern than all the shells I found" message should suggest something These require a deep delve into one or more less-popular and/or archaic systems. * #110324: autoupdate does some nonsensical changes * #110886: autoupdate produces broken configure.ac Signing up for either of these is tantamount to signing up to rewrite autoupdate from scratch. I believe that the only viable approach will be, as mentioned in 110886, an independent parser (not based on M4) specialized for machine-assisted refactoring of Autoconf scripts. If you don't see a path to fixing both of these *and* all of the autoupdate-related expected failures in the testsuite, all at once, probably best to leave bad enough alone. * #110325: autoupdate introduces a second invocation of AC_OUTPUT This one _could_ be fixed within the existing autoupdate _but_ you would need to add a feature to GNU M4 first. * #110448: AC_CONFIG_AUX_DIRS fails with Windows paths * #110480: Generated configure script has errors * #110573: AR_CONFIG_AUX_DIR(S) does not recognize shell vars nor absolute paths * #110852: autoconf 2.71 checking command to parse /usr/bin/nm -B output from gcc object These need clarification from the bug reporters and I'm pretty sure that at least two of them are not our bug at all. zw
