Signed-off-by: Steve Bennett <steveb@workware.net.au>

In list -> dict conversion, the new key needs to
be decremented.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

When converting a list to a dict, ensure that the last duplicate
wins, not the first duplicate.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

If a hash collision occurs and then the original
entry that cased the hash collision is removed, we
still need to continue iterating when searching for the
new key. Don't stop at the now-empty slow.
So mark these entries with offset=-1 to indicate that they
need to be skipped.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Although the documentation has always stated that, like Tcl,
insertion order of dictionaries was preserved, this has never
been the case. Instead, dictionaries were implemented as simple
hash tables that did not preserve order.

Now, a new implementation of dictionaries preserves insertion
order and has a number of other benefits.

Instead of hashing keys and storing keys and values in the hash table,
the keys and values are not stored in a separate table, exactly as
lists are stored, with alternating key, value pairs. Iterating over the
dictionary is exactly like iterating over a list, where the order is consistent.

The hash table uses closed hashing rather than open hashing to avoid
allocatation of hash entry structures. Instead a fixed (but expandable)
hash table maps the key hash to the offset in the key/value table.
This use of offsets means that if the key/value table grows, the offsets
remain valid. Likewise, if the hash table needs to grow, the key, value table
remains unchanged.

In addition to the offset (which indexes to the value, and 0 means the hash table entry is unused),
the original hash is stored in the hash table. This reduces the need for object
comparisons on hash entry collisions.

The collision resolution algorithm is the same as that used by Python:

	peturb >>= 5;
	idx = (5 * idx + 1 + peturb) & dict->sizemask;

In order to reduce collisions, the hash table is expanded once it reaches
half full. This is more conservative that Python where the table is expanded
when it is two thirds full.

In addition, the new faster hashing algorithm from Tcl 8.7 is used.
This the hash for integers to be calculated efficiently without requiring
them to be converted to string form first.

This implementation is modelled largely on the Python dict implementation.

Overall the performance should be an improvement over the current implementation,
whilst preserving order. Dictionary creating and insertion should be faster
as hash entries do not need to be allocated and resizing should be slightly faster.
Entry lookup should be about the same, except may be faster for pure integer keys.

Below are some indicative benchmarks.

                                                    OLD     NEW
dict-create-1.1  Create empty dict               97.2ns       .
dict-create-1.2  Create small dict                440ns    -27%
dict-create-1.3  Create medium dict              1.54us    -57%
dict-create-1.4  Create large dict (int keys)     130us    -80%
dict-create-1.5  Create large dict (string keys)  143us    -75%
   dict-set-1.1  Replace existing item            258ns    -34%
   dict-set-1.2  Replace nonexistent item         365ns    -49%
dict-exists-1.1  Find existing item              55.7ns     -5%
dict-exists-1.2  Find nonexistent item           55.0ns     -5%

Signed-off-by: Steve Bennett <steveb@workware.net.au>

We can simply use Jim_ReplaceHashEntry() when replacing a command
in non-local mode now that the proc epoch doesn't need to be incremented.

And JimRemoveHashEntry() only has one caller, so fold it into
Jim_DeleteHashEntry()

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Don't look up every object in the command table, since it can be slow
to do that. Only lookup if the object looks like a reference.

Also, script and expression objects can't contain references that
aren't already contained in sub-objects, so there is no need to
scan them for references.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

In this case, show the result as a decimal value

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Under some circumstances, such as lrepeat and lreverse we know
the length of the final list, so allocate the final size immediately
rather than growing the table in multiple steps.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Pass JIM_NONE to prevent an error message being generated
which is subsequently thrown away.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

The changes in cdfa4637

 broke TEST 35 because the cmd
pointer is dereferenced before checking the procEpoch instead
of afterwards.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Now argument is Jim_Obj *, not const char *

Signed-off-by: Steve Bennett <steveb@workware.net.au>

If running in a VM with very variable time

Signed-off-by: Steve Bennett <steveb@workware.net.au>

namespace, ssl

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Instead of incrementing the proc epoch on every command removal
and some command creation, cache previous deleted commands
(empty structure only). Periodically increment the proc epoch
and invalide all cached commands.

This is especially a win when creating short lived commands.
e.g.

	proc a {} {
		local proc b {} {
			# do something
		}
		# now b is removed
	}

Signed-off-by: Steve Bennett <steveb@workware.net.au>

The hash table used to store commands now uses Jim_Obj keys rather
than allocated char *, so embedded nulls are supported.

This means that some API function such as Jim_RenameCommand()
now take Jim_Obj * rather than const char *, however Jim_CreateCommand()
is retained with const char * for convenience and the new Jim_CreateCommandObj()
is added.

This is generally a performance win as the existing Jim_Obj can be used
as the key.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

string match, switch -glob, info commands, etc. all now support
patterns and strings with embedded nulls.

Fixes #143

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Hash tables used to store variables are now use Jim_Obj keys rather
than allocated char *, so embedded nulls are supported.

This is generally a performance win as the existing Jim_Obj can be used
as the key.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Signed-off-by: Steve Bennett <steveb@workware.net.au>

The fix for tailcall in 0d5a208e

 meant that a JIM_RETURN return
code from apply was no longer being converted to the appropriate
return code.o

Fixes: #157

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Fixes #155

Signed-off-by: Steve Bennett <steveb@workware.net.au>

An implicit concat was being done on the arguments to 'super', so
arguments containing spaces were mangled.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Using a custom "expect-like" module to allow testing Jim in interactive
mode. This also exercises the 'socket pty' support.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

If the pattern begins with -, the internal invocation of regexp
will treat the pattern as an option. Fix this by adding -- to the
internal invocation of regexp.

Fixes #154

Reported-by: Barry Arthur <barry.arthur@gmail.com>
Signed-off-by: Steve Bennett <steveb@workware.net.au>

Skip ipv6 tests if not supported

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Allows a psuedo-tty pair to be created.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Signed-off-by: Steve Bennett <steveb@workware.net.au>

For some SSL connections it is necessary to set the Server Name
Indication in the connection in order to receive the correct
certificate. Allow this as part of the client ssl call with:

  $sock ssl -sni $servername

Also for -server mode, allow the certificate and private key to be
stored in a single file and only be specified once.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

We can't use feof() and 'buffering none' on ssl connections.
Instead we have to get eof from the ssl layer, and provide
special handling for buffering in the eventloop.

For eof, add ssl_eof() and detect SSL_read() results that indicate
eof to set AIO_EOF in flags.

For buffering, add 'read -pending' that will read, and then immediately
read any buffered data so that the 'readable' event will trigger next
time.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Note that there is currently a problem with ssl and readable events

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Basic testing of each of the socket types

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Because we use catch -exit { ... }, if a test uses os.fork
we will return in both the parent in the child. To fix this,
require the child to use exit 99, and detect this case and exit
from the child in this case.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

If exit is caught with catch -exit, it is currently not possible
to retrieve the exit value. If an exit code is provided, set it
as the interp result.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

While -D__MINGW_USE_VC2005_COMPAT enables a 64-bit time_t,
it is necessary to explicitly use _stat64 to access the stat
structure containing 64 bit times.

Signed-off-by: Steve Bennett <steveb@workware.net.au>

Mostly just avoiding running certain tests, but also:
- rename() won't overwrite an existing file on Windows
- ensure that eof returns 0 or 1
- in aio.test, create and read the file in binary mode

Signed-off-by: Steve Bennett <steveb@workware.net.au>

In case the index is invalid, the string should contain -MAX_INT

Signed-off-by: Steve Bennett <steveb@workware.net.au>

numPids is always zero here so the code does nothing

Signed-off-by: Steve Bennett <steveb@workware.net.au>