Configure UX

[erlend-aasland]

Feature or enhancement

The LDLIBRARY and HOSTRUNNER checks overlap; this results in garbled output:

$ ./configure
[...]
checking LDLIBRARY... checking HOSTRUNNER... 
libpython$(VERSION)$(ABIFLAGS).a

The IPv6 library check is very subtle:

$ ./configure
[...]
configure: using libc

Suggesting to clean this up, for improved configure user experience:

$ ./configure
[...]
checking LDLIBRARY... libpython$(VERSION)$(ABIFLAGS).a
checking HOSTRUNNER...
[...]
checking ipv6 stack type... kame
checking ipv6 library... libc

Linked PRs

• gh-115806: Improve configure output #115807
• gh-115806: Don't mangle output of configure 'ipv6 stack type' check #116165
• gh-115806: Don't mangle gdbm header dependency detection output #129390

[erlend-aasland]

Ah, our PR gh-116016 mangled the output again:

checking ipv6 stack type... checking for IPV6_INRIA_VERSION defined in netinet/in.h... no
checking for __KAME__ defined in netinet/in.h... yes
kame

With #116165 applied, we get this:

checking for IPV6_INRIA_VERSION defined in netinet/in.h... no
checking for __KAME__ defined in netinet/in.h... yes
checking ipv6 stack type... kame
checking ipv6 library... libc

[erlend-aasland]

configure -C still produces some garbled output:

checking for library containing dbm_open... (cached) -lgdbm_compat
checking for ndbm presence and linker args... yes (-lgdbm_compat)
(cached) (cached) checking for library containing dbm_open... -lgdbm_compat

[hugovk]

Triage: the linked PRs are merged, can this be closed?

[encukou]

I'll close it now. If there are more improvements to do, please re-open or open a new issue.

Thank you for the improvements!

[erlend-aasland]

#115806 (comment) is still not fixed. The problem is that we're using AC_CHECK_HEADER wrapped in AC_CACHE_VAL:

cpython/configure.ac

Lines 4431 to 4454 in 8e57877

	dnl "gdbm-ndbm.h" and "gdbm/ndbm.h" are both normalized to "gdbm_ndbm_h"
	dnl unset ac_cv_header_gdbm_ndbm_h to prevent false positive cache hits.
	AS_UNSET([ac_cv_header_gdbm_ndbm_h])
	AC_CACHE_VAL([ac_cv_header_gdbm_slash_ndbm_h], [
	AC_CHECK_HEADER(
	[gdbm/ndbm.h],
	[ac_cv_header_gdbm_slash_ndbm_h=yes], [ac_cv_header_gdbm_slash_ndbm_h=no]
	)
	])
	AS_VAR_IF([ac_cv_header_gdbm_slash_ndbm_h], [yes], [
	AC_DEFINE([HAVE_GDBM_NDBM_H], [1], [Define to 1 if you have the <gdbm/ndbm.h> header file.])
	])
	AS_UNSET([ac_cv_header_gdbm_ndbm_h])
	AC_CACHE_VAL([ac_cv_header_gdbm_dash_ndbm_h], [
	AC_CHECK_HEADER(
	[gdbm-ndbm.h],
	[ac_cv_header_gdbm_dash_ndbm_h=yes], [ac_cv_header_gdbm_dash_ndbm_h=no]
	)
	])
	AS_VAR_IF([ac_cv_header_gdbm_dash_ndbm_h], [yes], [
	AC_DEFINE([HAVE_GDBM_DASH_NDBM_H], [1], [Define to 1 if you have the <gdbm-ndbm.h> header file.])
	])
	AS_UNSET([ac_cv_header_gdbm_ndbm_h])

I wonder if we're better off by using AC_CACHE_CHECK and AC_PREPROC_IFELSE for these.