5.4 open issues

From Net-SNMP Wiki
Jump to: navigation, search

We try to collect significant open issues for the 5.4 release cycle here. Please add your contributions/comments and fix as much as you can!

Open issues for 5.4.1

Issue Bug/patch Comment Assigned to
snmptrap/trapsess send to port 161 Patch 1620424 snmptrap and trapsess send to port 161 instead of port 162 by default. Patch from magfr available.  ?
error on subcontainer 'ia_addr' insert (-1) Bug 1693039, Patch 1712645 Provide a more meaningful log message for duplicate IP addresses instead of spamming the log continuously. rstory?
OSX embedded perl broken Bug 1377478 Embedded perl (and sometimes even the perl modules) seem to fail silently (and for unknown reasons) on OSX for some people, but not all.
Too vague a problem description to address for 5.4
 ?
Linux CPU stats Bug 1600261 Broken for Montavista Linux on XScale/ARM and possibly others. See bug description. Regression over 5.3.x.
Appears to be very hardware-specific. Not realistic to address for 5.4
 ?
64bit nlist Bugs 148532, 1484814, 1388562 nlist fails on 64bit systems (AIX, IRIX, Tru64). We need 64bit capable auto_nlist replacement routines. Every system has its own variant and changes are needed all over the net-snmp code. Stop-gap fix is specifying --without-kmem-usage which is now possible since 5.4.rc2. tanders
Solaris perl<->host issues (due to largefiles) Bug 1486366 Solaris system-supplied Perl is compiled with largefile support. The very same options break the build of the host module, though.
(Wes: This will not be fixed by 5.4.) At least detected (and embedded perl disabled) in pre3.
 ?
ifTable transient interface handling Bug 1238981, Patch 1513191 offer an (per-interface) config option how transient interfaces are handled in ifTable (same or different). A List FAQ.
This feels like a new feature, and not appropriate for 5.4
rstory?
disman/event crash Bug 1556259 Reported crashes in mteTrigger.c
This is a race condition between (typically high-volume) request handling and trap generation. Not realistic to address for 5.4
dts12
Suppress error messages on SIGHUP  ? Remove the likely harmless, but annoying error messages on SIGHUP. A List FAQ
Handled for "extend" and "monitor" directives. What others?
rstory?/dts12
library layering Bug 1565496 (OSX), Patches 1398873 1493522 Consensus seems to be to extend --enable-as-needed for 5.4.1 rather than 5.4 and come up with a proper configure test for 5.5. --Thomas. Full analysis / Work in progress hardaker/tanders
fix perl compiler warnings - There are a number of nasty compiler warnings for the perl modules that needs to be dealt with. List provided to hardaker. (http://pastebin.ca/196857) Dave has flagged three issues as potentially important after review (see list). Full fixes for SNMP.pm are in 5.4.x CVS. tanders
OSX CPU stats Patch 1600522 Neither cpu_sysctl (preferrable?) nor cpu_nlist compile on Darwin. rc1/rc2 ship cpu_null for Darwin. To be investigated! At least it doesn't seem to be a regression over 5.3.x, based on rstory's snmpwalk output for darwin7/8. <P/>A new cpu_mach.c has been supplied as a patch after rc3!
Too late for 5.4. Now in 5.4.x CVS.
dts12/rstory/jalbert


Fixed for 5.4

The following issues have been identified and fixed during the 5.4 release cycle.

Issue Bug/patch Comment Assigned to
man page aliases - The man page aliases are regenerated at each make invocation. Fixed for 5.4 final.(?) hardaker
MSVC workspace build - RFV in progress to fix Win32 MSVC Workspace build. Fixed for 5.4 final.(?) Alex_B
net-snmp.spec - dist/net-snmp.spec simply doesn't work as-is. Fixed for 5.4 final. tanders
snmplib/system.c #elif Bug 1598927 RFV in progress to fix #elif bug in snmplib/system.c as introduced by rc3. Fixed for rc4/final. tanders
snmptrapd -A Bug 1587759 RFV in progress to add "-A" to list of valid snmptrapd options (implementation already there). Fixed for rc4/final. tanders
UCD memory inconsistencies II - RFV in progress to address inconsistencies wrt. memTotalFree and memSwapError{,Msg}. To be considered for rc3. Currently at +2 for Dave's original patch.
Proposal withdrawn. Fixed MIB text only for 5.4 final.
dts
missing IP scalars in snmpwalk Bug 1554989 snmpwalk doesn't show IP scalars on all platforms!
Doesn't affect 5.4 - just 5.[23].x lines
rstory?
OpenBSD4 system header - See 9/26 and 9/27 irc logs; openbsd4 uses openbsd.h, and thus a lot of openbsd3 code that most likely works on openbsd4 is excluded.
Some fixes are in pre3. Rework openbsd system headers and defines?
Current openbsd4 (released Nov 1) support seems to be working, based on build testing feedback from Grey on IRC.
 ?
notifications over TCP broken Bug 1571945 The recent patch for the UDP/IPv4 response IP address broke other transports. See the bug report for a decent analysis.
Fix for the immediate problem applied - a cleaner solution can probably wait
 ?
UCD memory inconsistencies I - Should 'memAvailSwapTXT' and 'memAvailRealTXT' report the amount of memory free, or the amount used? The MIB object name implies one, the description the other! Inconsistently implemented in 5.3.x. Implemented as "free" in 5.4.x.
Fix to revert to reporting the amount used, document the problem and add replacement (less confusingly defined) MIB objects done for rc2.
dts
sed issue II - On Solaris 2.6/7 "configure --with-mib-modules=if-mib,udp-mib,tcp-mib" yields into an empty Makefile. Looks like the previous sed issue, but now the "@mibgroup_list_o" list is getting too long. Wes has created a fix and is now at +3 at least. Fixed for rc2. hardaker
proxy tests fail for TCP - "make test TESTOPTS='-P tcp'" fails for the proxy GET/SET tests. Skip proxy tcp tests in rc2. tanders
UCD memCached.0 - RFV in progress to re-add memCached.0 on Linux/FreeBSD. Fixed for rc3. tanders
configure perlcc detection broken - RFV (2 character patch) posted to fix perlcc detection. Fixed for rc3. tanders
add missing manual pages - Six manual pages that were missing have recently been committed. RFV accepted to add them to the build process. Fixed for rc3. tanders
nlist build issues - RFV in progress to properly protect the nlist code in snmplib/system.c to allow for --without-kmem-usage (e.g. on IRIX). Fixed for rc3. tanders
FreeBSD cpu_nlist hang - 5.4.rc1 on SF CF machine x86-freebsd1 (FreeBSD 5.4) hangs badly in cpu_nlist initialization, even on a simple "snmpd -H". Fixed for rc3. dts12
SNMP.pm MIB code broken on 64-bit - "make perltest" fails on all 64-bit platforms in t/mib.t. Fixed for rc3. tanders
snmptrapd partial oid matching Patch 1536773 introduce partial oid matching for traphandle, e.g. "traphandle someOID* /call/on/subtree/matching" (completely rewritten by Wes and now requires a -t flag).
It's in pre1. Changed to "oid[.]*" for pre2.
hardaker
no perl/agent/Support/Makefile - With some perl versions "make perlmodules" will not create perl/agent/Support/Makefile from Makefile.PL, so the build will fail (e.g. on Solaris 8 with system-supplied Perl).
Fixed in pre1.
hardaker/tanders
pre/rc version numbers - 5.4.pre1 gets rewritten to 5.4.0.1 in net-snmp-config which breaks the perl version checks and the default build. See -coders.
Fixed for pre2.
hardaker/tanders
Windows build broken - 5.4.pre1 doesn't build on Windows (as reported by Alex) due the recent changes to the transport code (patch 1282566).
Fixed for pre2.
Alex_B
make perltest - Currently fails on perl/SNMP t/notify test (tests 7, 9, 10). It should pass perfectly in order to be able to use it for regression testing.
Fixed for pre2.
tanders
python binding - Initial python binding (just client-side get/set/walk etc.).
Added before pre2, including a --with-python-modules.
hardaker/jmarzot
smarter "enable embedded perl by default" - Only enable embedded perl if it will successfully compile later on.
Quite smart in pre3, but some problems (minimum perl version?) still need to be detected. (we don't know the minimum perl version... if we know it, we can test for it. It should be quite old though (I believe it works under 5.004)
hardaker/tanders
Solaris 5.9 build broken Bug 1569150 -Usolaris2 -Dsolaris2=solaris2 doesn't seem to work properly.
Fixed for pre4.
dts12
UCD-SNMP-MIB::systemStats - Reported to be broken on some systems (Solaris, some RHEL Linux machines). Breaks "make test" (CPU test on Solaris, snmpwalk test on Linux).
Solaris and Linux parts fixed for pre4.
dts12
OpenBSD/FreeBSD/Solaris if-mib rewrites - the suggestion in the NEWS file, --with-mib-modules=if-mib, doesn't work. Need a way to have configure pass NETSNMP_INCLUDE_IFTABLE_REWRITES during module processing/compilation. Add --enable-mfd-rewrites (see updated NEWS).
Fixed for pre4
rstory
HOST-RESOURCES-MIB::hrStorageIndex - Indexes have been re-ordered from previous versions. Have we justified this/notified users of this change?
Documented in NEWS
dts12/bshaw
Solaris 2.6/7 configure sed failures - On Solaris 2.6 and 7 configure reproducibly fails with "Output too long" and "sed: garbled command" errors from sed during configure's AC_OUTPUT for mib_module_list_h. This list is extremely long (>6500 chars on my Linux build) causing trouble on many systems :-(.
Fixed properly using files instead of sed -- Wes
hardaker


Documentation and testing work for 5.4 final

Item Comment Assigned to
manual pages quite a few manual pages currently ship with explicit XXX - to be filled in for 5.4 final.  ?
missing tests We're currently missing tests for at least: disman/event monitoring, snmpvacm extensions, snmptranslate, snmpusm DH stuff, snmptrapd walk, ...  ?
README.smux Very old. Updated for 5.4 final. tanders/sfriedl
README.solaris Outdated in many regards, e.g. the complains on disman/event. Improved for 5.4 final. tanders