Unbound 1.9.3 released
yvoinov at gmail.com
Tue Aug 27 16:31:50 UTC 2019
Same shame, Wouter:
/usr/local/lib/libevent.so -lsendfile -lrt -lsocket -lnsl -lcrypto -lmd
-pthreads -mt -R/usr/local/lib -R/usr/local/lib -R/opt/csw/lib
gmake: *** [Makefile:335: unbound-control] Error 2
ld: fatal: soname option (-h, --soname) is incompatible with building a
ld: fatal: flags processing errors
postopt: error: ld failed to link the binary
Does not help.
27.08.2019 21:24, Wouter Wijngaards via Unbound-users пишет:
> Hi Yuri,
> Another option, if your system has a specific version of libtool for it,
> is to specify the libtool version specificaly on the configure line,
> like this ./configure libtool=/usr/bin/libtool (or point to another
> location or version of libtool). Then it uses that libtool for dynamic
> library linking, and if that passes the right options, things could
> start to work.
> Best regards, Wouter
> On 8/27/19 5:17 PM, Wouter Wijngaards via Unbound-users wrote:
>> Hi Yuri,
>> If that is any similar to yours, the ./configure --disable-shared flag
>> could fix it.
>> That stops unbound from building the libunbound.so dynamic library. The
>> static lib is still built, and could be used if you wanted to. The
>> unbound server binary and commandline tools are then built when it
>> continues the build process. It also links like normal with dynamic
>> libraries, so I think it should be (relatively) harmless to add to the
>> configure options.
>> Or, from the link you cite, switch to either gcc (with its ld), or to
>> solaris 11. The commandline switches that it complains about, are
>> created by libtool, so, instead, upgradeing to a different version of
>> libtool, installing that, autoreconf, and then configure may result in a
>> different build process (if that libtool supports the build system).
>> Best regards, Wouter
>> On 8/27/19 4:59 PM, Yuri via Unbound-users wrote:
>>> 1.9.3 failed to link on Solaris with Oracle Developer Studio with this
>>> ld: fatal: soname option (-h, --soname) is incompatible with building a
>>> dynamic executable
>>> ld: fatal: flags processing errors
>>> It seems like this issue: https://gitlab.haskell.org/ghc/ghc/issues/4973
>>> Is is possible to fix?
>>> 27.08.2019 14:22, Wouter Wijngaards via Unbound-users пишет:
>>>> Unbound 1.9.3 is available:
>>>> sha256 1b55dd9170e4bfb327fb644de7bbf7f0541701149dff3adf1b63ffa785f16dfa
>>>> pgp https://nlnetlabs.nl/downloads/unbound/unbound-1.9.3.tar.gz.asc
>>>> This release has a number of bug fixes. Added is the ipset module, that
>>>> helps add ip-addresses that are looked up in a domain to a firewall
>>>> ip-address filter. Also, the python module has restart next, per-query
>>>> data and multiple instance support. The unbound -V option has been
>>>> added and it prints the build config.
>>>> - PR #28: IPSet module, by Kevin Chou. Created a module to support
>>>> the ipset that could add the domain's ip to a list easily.
>>>> Needs libmnl, and --enable-ipset and config it, doc/README.ipset.md.
>>>> - Merge PR #6: Python module: support multiple instances
>>>> - Merge PR #5: Python module: define constant MODULE_RESTART_NEXT
>>>> - Merge PR #4: Python module: assign something useful to the
>>>> per-query data store 'qdata'
>>>> - Introduce `-V` option to print the version number and build options.
>>>> Previously reported build options like linked libs and linked modules
>>>> are now moved from `-h` to `-V` as well for consistency.
>>>> - PACKAGE_BUGREPORT now also includes link to GitHub issues.
>>>> Bug Fixes:
>>>> - Fix #39: In libunbound, leftover logfile is close()d unpredictably.
>>>> - Fix for #24: Fix abort due to scan of auth zone masters using old
>>>> address from previous scan.
>>>> - Fix to omit RRSIGs from addition to the ipset.
>>>> - Fix to make unbound-control with ipset, remove unused variable,
>>>> use unsigned type because of comparison, and assign null instead
>>>> of compare with it. Remade lex and yacc output.
>>>> - make depend
>>>> - Added documentation to the ipset files (for doxygen output).
>>>> - Fix python dict reference and double free in config.
>>>> - Fix memleak in unit test, reported from the clang 8.0 static analyzer.
>>>> - For #45, check that 127.0.0.1 and ::1 are not used in unbound.conf
>>>> when do-not-query-localhost is turned on, or at default on,
>>>> unbound-checkconf prints a warning if it is found in forward-addr or
>>>> stub-addr statements.
>>>> - Fix for possible assertion failure when answering respip CNAME from
>>>> - Fix in respip addrtree selection. Absence of addr_tree_init_parents()
>>>> call made it impossible to go up the tree when the matching netmask is
>>>> too specific.
>>>> - Fix #48: Unbound returns additional records on NODATA response,
>>>> if minimal-responses is enabled, also the additional for negative
>>>> responses is removed.
>>>> - Fix #49: Set no renegotiation on the SSL context to stop client
>>>> session renegotiation.
>>>> - Fix question section mismatch in local zone redirect.
>>>> - Add verbose log message when auth zone file is written, at level 4.
>>>> - Add hex print of trust anchor pointer to trust anchor file temp
>>>> name to make it unique, for libunbound created multiple contexts.
>>>> - For #52 #53, second context does not close logfile override.
>>>> - Fix #52 #53, fix for example fail program.
>>>> - Fix to return after failed auth zone http chunk write.
>>>> - Fix to remove unused test for task_probe existance.
>>>> - Fix to timeval_add for remaining second in microseconds.
>>>> - Check repinfo in worker_handle_request, if null, drop it.
>>>> - Generate configlexer with newer flex.
>>>> - Fix warning for unused variable for compilation without systemd.
>>>> - Fix #59, when compiled with systemd support check that we can properly
>>>> communicate with systemd through the `NOTIFY_SOCKET`.
>>>> - iana portlist updated.
>>>> - Fix autotrust temp file uniqueness windows compile.
>>>> - avoid warning about upcast on 32bit systems for autotrust.
>>>> - escape commandline contents for -V.
>>>> - Fix character buffer size in ub_ctx_hosts.
>>>> - Option -V prints if TCP fastopen is available.
>>>> - Fix unittest valgrind false positive uninitialised value report,
>>>> where if gcc 9.1.1 uses -O2 (but not -O1) then valgrind 3.15.0
>>>> issues an uninitialised value for the token buffer at the str2wire.c
>>>> rrinternal_get_owner() strcmp with the '@' value. Rewritten to use
>>>> straight character comparisons removes the false positive. Also
>>>> valgrinds --expensive-definedness-checks=yes can stop this false
>>>> - Please doxygen's parser for "@" occurrence in doxygen comment.
>>>> - Fixup contrib/fastrpz.patch
>>>> - Remove warning about unknown cast-function-type warning pragma.
>>>> - Document limitation of pidfile removal outside of chroot directory.
>>>> - Fix log_dns_msg to log irrespective of minimal responses config.
>>>> - Fix that pkg-config is setup before --enable-systemd needs it.
>>>> Best regards, Wouter
"C++ seems like a language suitable for firing other people's legs."
* C++20 : Bug to the future *
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 659 bytes
Desc: OpenPGP digital signature
More information about the Unbound-users