[nmbro]

Fixes #3781, found while researching which platforms still need libusb-0.1 for #1507.

Problem

• configure required libusb_strerror() (libusb 1.0.16+, Add autoconf check for libusb_strerror() #509) to use the libusb-1.0 API, so e.g. the libusb1 1.0.9 package of CentOS 6.3-6.10 was rejected.
• With --with-usb=auto (or yes), a rejected libusb-1.0 also disabled USB drivers entirely, without trying an available libusb-0.1: the 0.1 checks only ran when pkg-config did not know libusb-1.0 at all.

Building current master against libusb 1.0.9 with that check bypassed showed only two methods missing: libusb_strerror() and libusb_get_version().

Changes

1. build: accept libusb-1.0 from 1.0.9 on
   • configure: accept libusb_error_name() (1.0.9+) when libusb_strerror() is missing; check for libusb_get_version(). Both the detection and the cached pass.
   • usb-common.h: without libusb_strerror(), map it to libusb_error_name(). Log messages then show the code name (e.g. LIBUSB_ERROR_TIMEOUT) instead of a description.
   • libusb1.c: without libusb_get_version(), driver.version.usb reports libusb-1.0 (version unknown, 1.0.9 or older).
   • nut-scanner: if the libusb loaded at run time lacks libusb_strerror(), use libusb_error_name(); a missing libusb_get_version() is treated as an unknown (old) version instead of failing to load the library. struct libusb_version is mirrored (stable ABI) for builds against headers that lack it, so a newer run-time libusb can still be asked.
2. build: fall back to libusb-0.1 if libusb-1.0 is too old
   • When the libusb-1.0 picked by preference does not link with libusb_init() and libusb_error_name(), and a libusb-0.1 was also detected, switch to libusb-0.1 before computing flags, with a notice. Explicit --with-usb=libusb-1.0 or custom --with-usb-includes/--with-usb-libs still fail as before. Replaces an existing TODO comment about this case.

Testing

Built against headers and libraries extracted from the CentOS 6.0/6.10 RPMs (PKG_CONFIG_LIBDIR/PKG_CONFIG_SYSROOT_DIR, host GCC; the centos:6 container itself does not run on a current kernel without vsyscall=emulate):

libusb installed	--with-usb=	Result
1.0.9	auto	libusb-1.0; all USB drivers and nut-scanner build, no new warnings
1.0.9 + 0.1.12	auto	keeps libusb-1.0
1.0.3 + 0.1.12	auto	falls back to libusb-0.1; all USB drivers build
1.0.3	libusb-1.0	fails: neither libusb_strerror() nor libusb_error_name()
host 1.0.27	auto, --enable-Werror	unchanged; drivers and nut-scanner build clean
host 1.0.27 + EL6 0.1.12	auto	probe passes, keeps 1.0.27

Also: cached configure -C re-run of the fallback case; the first commit alone against 1.0.9; nut-scanner built against 1.0.27 run with the 1.0.9 library (both fallbacks taken, library loads, scan runs) and vice versa; make distcheck-light; make spellcheck. Not tested with USB devices on libusb 1.0.9.

libusb 1.0.3 (CentOS 6.0-6.2 only; a yum update brings 1.0.9) would additionally need a fallback for libusb_error_name(), libusb_handle_events_timeout() for the _completed variant (used by apcmicrolink), and an empty LIBUSB_CALL; I left that out.

Checklist

• Described the changes in the PR submission
• Multiple commits, each with a functional change
• Use of coding helper tools and AI disclosed: Claude Code was used for development assistance
• Revised for NUT code style and ASCII-only text; configure probes are guarded, no new pragmas
• make distcheck-light passes
• Added bullet points into NEWS.adoc
• Passed make spellcheck

🤖 Generated with Claude Code

[github-actions]

A ZIP file with standard source tarball and another tarball with pre-built docs for commit 8ca85cf is temporarily available: NUT-tarballs-PR-3782.zip.

[jimklimov]

Looks superb, thanks!

[AppVeyorBot]

✅ Build nut 2.8.5.5921-master completed (commit 14a50ef4fb by @nmbro)

• artifacts
• A 7-zip archive with built NUT for Windows debug binaries is temporarily available: NUT-for-Windows-x86_64-SNAPSHOT-2.8.5.5921-master.7z
• ccache: 30 of 577 compilations were cache hits (5.2%); cache size 749 of 750 MiB after cleanup