macLib failes to suppress all warnings when its FLAG_SUPPRESS_WARNINGS flag is set. Ben Franksen posted a patch for this problem.
These problems are known to exist in all R3.14.8.x releases:
These problems are known about in the R3.14.8.2 release:
Sites that want to make use of the EPICS_SITE_VERSION setting in CONFIG_SITE will have to apply this patch in the configure directory to get their site version information to appear in the epicsVersion.h generated file.
Since the calcoutRecord had asynchronous device support written for it in Base R3.14.5, it has been susceptible to infinite FLNK loops that will crash the IOC if they occur. The fix is simple.
The fix for this problem has been committed and will be in the next base release. A workaround for all non Visual C++ 2005 builds on WIN32 is to edit "$<*" to "$<" in configure/RULES_BUILD. Workarounds for Visual C++ 2005 builds are to use a cygwin version of GNU make, or copy the files configure/RULES_BUILD, configure/os/CONFIG.win32-x86.win32-x86, and configure/tools/useManifestTool.pl from the latest nightly R3.14.x CVS snapshot file.
Reported problems with the epicsTime implementation (See Mantis #245 for details) can be fixed by applying this patch to the file src/libCom/osi/epicsTime.cpp.
The input record types ai, bi, mbbi and mbbiDirect that support SIMM=RAW are subject to a bug which causes these records to go into INVALID alarm whenever SIMM is true, ignoring the severity set in the SIMS field. This can be fixed by applying this patch inside base/src/rec.
This problem was fixed in the R3.14.8.2 patch release:
Architectures based on Posix may have problems with a recursive definition for epicsStrtod which was introduced during the Win32 cleanup process. This patch should be applied in the src/libCom/misc directory to resolve the problem. We may generate a new patch release containing this fix in a few days, although the final resolution may be slightly different to this patch.
This problem exists in R3.14.8.1 only, and has been seen on Linux and MacOS/Darwin; Solaris may be affected as well.
These problems were fixed in the R3.14.8.1 patch release:
There is a scheduling vulnerability in the CA client code which can cause delays in recognizing IOC disconnections and possibly slow connection times for some kinds of CA client applications. This patch file contains a fix for this problem, which is filed in the Mantis bug tracker as Issue #232. Apply the patch in the src/ca directory.
A number of IOC symbols do not contain the correct epicsShare... magic keywords needed to use them on an MS Win32 system. These corrections are being made in CVS as they are found.
The fix for this is to replace the VX_READ in the vxDevWriteProbe() routine (which is in base/src/libCom/osi/os/vxWorks/devLibOSD.c) with VX_WRITE.