The select() implementation utilizes its own set size limit
that should be checked.

When calculating the relative time interval for the select() call,
the increment should have been a decrement.

This also adds a UT case that specifically exercises the carryover
described.

The only way for this test happen would be if somehow the normal
unlock process was bypassed.

The return statement from OS_FileSys_FindVirtMountPoint() was
performing several match operations and was hard to understand.

This breaks up the statement so it is easier to read and adds
some informational comments.

The "U32" designator in the name had become confusing because all
the members had over time migrated to a non-uint32/dedicated type.

The point is to pass a value through a void* so a name change better
indicates that purpose.

This is documented as a possible errno from select on
some systems, and the call should be repeated in that case.

The ESPIPE errno means that seeking is not supported on the
given file handle.  Translate to the OS_ERR_OPERATION_NOT_SUPPORTED
error rather than not implemented as it better indicates the
actual condition.

Propagate a VxWorks error on moduleInfoGet() call into OS_ERROR return.

OSAL provides an abstract buffer for socket addresses, independent
of the underlying implementation.  The size of this buffer is
configurable by the user via compile-time options.

This adds a CompileTimeAssert to confirm that the size of this
abstract buffer is large enough to store any of the enabled
address types. This also removes the need for runtime tests.

A select() is used after connect() to provide application-controlled
timeout on the connection initiation.

This just adds some comments to describe why this is done.  It also
adds a debug statement if the connect fails for a reason other than
EINPROGRESS.

In VxWorks the impl calls timer_settime followed by timer_gettime on the
same timer to check if rounding occurred.  If the second call fails this
reports it as an error.

Note unless there is some sort of OS bug this should be impossible to
happen as this code only runs after a successful timer_settime.

Fix #875, Move copyblock size to define

Fix #873, Remove rogue while loop in OS_DeleteAllObjects

Fix #859, Consolidate duplicated switch in OS_SocketOpen_Impl

Fix #868, Use quotes for local includes

Fix #854, Const correct input pointers

Fix #851, Remove duplicate network prototypes

Fix #842, Remove NULL redefine

…_guide

Fix #911, Add Contributing Guide

…tests

Fix #752, Utilize UTASSERT_CASETYPE_NA to report OS_ERR_NOT_IMPLEMENTED

Fix #855, check against FD_SETSIZE in bsd-select

The select() implementation utilizes its own set size limit
that should be checked.

Fix #857, correct interval calculation in DoSelect

When calculating the relative time interval for the select() call,
the increment should have been a decrement.

This also adds a UT case that specifically exercises the carryover
described.

Fix #862, comments describing select after connect

Fix #858, add check for EAGAIN in addition to EINTR

Fix #861, compile time assert for sockaddr size

Fix #416, add shell functional test

Fix #867, better error translation for ESPIPE errno

Fix #869, rename OS_U32ValueWrapper_t

Fix #876, break up logic in return statement

Fix #889, report timer_gettime error

Fix #886, return moduleInfoGet error

Fix #882, make module comment same as other services

Fix #883, remove unreachable test

Add an extra limit check for the index, as it is possible
due to padding that this goes beyond the end of the array.

Fix #919, check index inside fdset conversions

Do not silently ignore a filehandle which was included in the OS_FdSet
but the "selectable" flag is false.  Instead translate this to
the OS_ERR_OPERATION_NOT_SUPPORTED error.

Fix #921, make non-selectable FD an error