The replication update process and its helper processes record statistics using the STATSDB facility for reporting using ^%YGBLSTAT. GT.M reports the WRL, PRG, WFL, and WHE statistics as part of the YGBLSTAT output. See the GT.M Administration & Operations Guide for more details. Previously, the replication update process and its helper processes did not record statistics for STATSDB. (GTM-F132372)

By default, GT.M proactively splits blocks when it detects significant restarts in an attempt to reduce overall restarts. MUPIP SET -FILE mumps.dat -PROBLKSPLIT=N where N is 0 disables proactive splitting, as do very large values of N. Values of N greater than 0 adjust the lower limit for the number of records below which GT.M does not consider splitting a block. While this is behavior is aimed at reducing restarts, because it reduces data density, it may also increase the number of blocks and even the tree depth. (GTM-F135414)

$ZJOBEXAM() recognizes an optional second argument of an expr that evaluates to a string as described for the argument of ZSHOW specifying one or more codes determining the nature of the information produced by the function. If the argument is missing or empty, GT.M operates as if it was a "*" and produces the associated context. This provides a way to suppress content that might contain PNI. Previously, $ZJOBEXAM() always produced a full context. (GTM-F135292)

$ZMALL[OCLIM] provides a way for a process to limit its process memory. When the value is zero (0), GT.M imposes no limit, although the OS still does. A positive value specifies a byte limit with a minimum of 2.5MB. A value of minus one (-1) provides a value of half the system imposed limit if any. When a request for additional memory exceeds the limit, GT.M does the expansion and then produces a trappable MALLOCCRIT warning. By default, some later request for memory is likely to produce a fatal MEMORY error, unless subsequent to MALLOCCRIT, a limit has been reestablished by SET ZMALLOCLIM to the same or higher limit, but one not exceeding any system limit. Note that GT.M allocates memory from the OS in large blocks so the interaction of $ZMALLOCLIM with memory growth is not exact. In the case of a MEMORY error, GT.M makes an attempt to marshal available memory to enable as graceful a termination as possible. Note that independent of this mechanism, the OS may kill the process without recourse if it determines the greed of the process for memory jeopardizes the viability of the system. When the integer byte value specified in a SET $ZMALLOCLIN=intexpr or, at process startup, by the $gtm_malloc_limit environment variable, specifies a positive value, GT.M uses the smaller of that value and any OS defined amount for the value of $ZMALLOCLIM. GT.M does not give errors or messages about its choice for $ZMALLOCLIM between a specified value and some other more appropriate value, so if the application needs to verify the outcome, it should examine the resulting ISV value. MEMORY errors are fatal and terminate the process. Previously, fatal MEMORY errors were not preceded by a trappable MALLOCCRIT warning. (GTM-F135393)

Operations that replace non-graphic characters with $[Z]CHAR() representations give a MAXSTRLEN error when the result would exceed 1MiB. Previously such operations ensured they had adequate space for the result, but could pass strings to other operations that might have no provision for dealing with strings longer than 1MiB. (GTM-DE201378)

USE SOCKET handles the ATTACH, DETACH, CONNECT, SOCKET and ZLISTEN deviceparameters appropriately; previously, certain arguments for these deviceparameters could cause a segmentation violation(SIG-11). (GTM-DE201380)

Address issues that can result in a segmentation violation (SIG-11) or similar failure.

GT.M appropriately detects divide by zero; previously there were some unusual combinations of calculations which produced a SIGINTDIV (SIG-8), which could not be trapped and therefore terminated the process.

The GT.M compiler appropriately handles optimization of literal FALSE post conditionals; previously these could cause segmentation violations (SIG-11). The GT.M compiler deals with argument-less BREAK and NEW commands combined with certain patterns of local variable use, and also with NEW commands having indirect arguments after a FOR command. Previously these could cause a segmentation violation (SIG-11).

The VIEW command for "NOISOLATION" handles various edge cases and reports a VIEWARGTOOLONG error when a list of global names exceeds 1024 bytes. Previously, some malformed global names could cause a segmentation violation (SIG-11) and using a string over 1024 bytes would cause a GTM assertion failure.

The ZSHOW command operates as expected when directing the output to a MUMPS local. Previously repeatedly running the ZSHOW command in an indefinite loop with the output directed to a MUMPS local could result in a segmentation violation (SIG-11).

The $FNUMBER() and $JUSTIFY() functions work as expected. Previously, results requiring long lengths could result in a segmentation violation (SIG-11).

The GT.M compiler protects against an indirect argument combined with a missing closing parenthesis in a $INCREMENT() function; previously some syntax with that characteristic produced a segmentation violation (SIG-11).

$ZSYSLOG() ignores Format Ascii Output (FAO) directives. Previously, $ZSYSLOG() did not ignore such directives resulting in a segmentation violation (SIG-11).

The USE command works as expected when the indirect device name contains an ill formatted string. Previously certain strings could cause a segmentation violation (SIG-11).

GT.M ignores unknown device class mnemonic spaces. Previously, if a mnemonic space other than "SOCKET" or "PIPE" was used, a subsequent OPEN resulted in a segmentation violation (SIG-11).(GTM-DE201386)

$[Z]CHAR(1E48) and similar numeric literal overflows produce a NUMOFLOW error at compile time; previously an optimization inappropriately treated this as $[Z]CHAR(0). (GTM-DE201388)

GT.M pattern match reports a PATMAXLEN error for a deeply nested pattern containing many indefinite ranges (. symbols). While we think it unlikely, it is possible that existing complex patterns may now report this error. Previously such a deeply nested pattern terminated with a segmentation violation (SIG-11). (GTM-DE201389)

CTRAP characters other than CTRL-x, where x corresponds to the non-graphic characters represented by $CHAR(n) with n 0 to 31 inclusive, have no effect, and do not show up in ZSHOW "D" output. On the keyboard those characters correspond to <CTRL-@> to CTRL-_> with the bulk of the range being <CTRL-A> to <CTRL-Z>. Note that CTRAP=$CHAR(3) has a different semantic than the other CTRAP characters, in that, when enabled and $PRINCIPAL is a terminal device type, <CTRL-C> can interrupt at any time, including when $PRINCIPAL'=$IO. GT.M only recognizes the other CTRAP characters when they appear in input to a READ command. Previously, characters outside the range with codes 0-31 showed up in ZSHOW "D" output with a modulo 32 $CHAR() representation and caused corresponding CTRAP recognition. (GTM-DE201390)

Indirection of the form @x@y ignores any comment at the end of the value of x; previously such a comment stopped the evaluation so the subscripts after the second @ were ignored. In addition, if the first (x) expression evaluates to the empty string, GT.M produces a VAREXPECTED error; previously that scenario produced incorrect results, sub-optimal errors or, in the worst case damage to an item at the top of the heap. Also, if the second (y) expression is the empty string, it no longer has the potential to lead to incorrect results or sub-optimal error reports. (GTM-DE201393)

SET $ZTRAP=$ETRAP works correctly; previously it tended to result in $ZTRAP="" regardless of the prior value of $ETRAP, but could also produce garbage values, while less likely, other assignments to $ZTRAP could also go awry. The workaround was to NEW $ETRAP, assign a the value of $ETRAP to a temporary, the empty string to $ETRAP and $ZTRAP to the value of the temporary. (GTM-F135433)

When the Database segment to file mapping uses an environment variable, and the environment variable is not defined, MUPIP BACKUP uses the environment variable name itself in constructing the file name, as does MUPIP CREATE. Previously MUPIP CREATE created database with undefined environment variable name as the database name, but MUPIP BACKUP failed to backup that existing database. (GTM-DE201305)

MUPIP LOAD -FORMAT=BINARY checks the record length against the data for a record; starting with V6.0-000, GT.M defines record length as the data in a node, but the utility in question still inappropriately included the key in its length check. (GTM-DE201381)

When MUPIP BACKUP arguments specify a list, the utility processes regions in the listed order, or, for names expanded by wildcard ("*"), alphabetically. Previously, MUPIP BACKUP ignored any user-specified order of regions, and processed regions in FTOK order, which tends to change with changes in operational conditions within the underlying file system.(GTM-F135842)

The configure script checks for and removes semsstat2, ftok, and getuid files while installing GT.M in an existing directory where a GT.M installation already exists. Note that GT.M versions starting from V6.3-014 no longer require these files. Previously, the configure script did not remove these obsolete executable files.(GTM-DE201825)

The Source Server defers a TLS renegotiation when a prior TLS renegotiation is pending. Previously, the Source Server logged the REPLWARN message, closed the replication connection, and attempted to re-establish the replication connection with the Receiver Server.(GTM-F135415)

GT.M provides protection from asynchronous events during routine management. Previously, asynchronous events like MUPIP STOP at inopportune times could cause a segmentation violation (SIG-11). This was only seen in development and has never been reported by a customer.(GTM-DE201392)

GT.M detects and recovers from certain integrity errors in the auto-relink control structures. GT.M also issues RLNKRECNFL and RLNKINTEGINFO messages to the system log to report an integrity check. Previously, GT.M terminated with an segmentation violation (SIG-11) or a GT.M assert when it encountered such integrity errors. (GTM-F135435)

GT.M handles SIGHUP appropriately when $PRINCIPAL has HUPENABLE set; in V7.0-001 due to changes in deferred event handling, error handing could encounter a GTMASSERT2. In addition, a TERMHANGUP error implicitly sets the device to NOHUPENABLE, so should a process anticipate multiple disconnects/hangups, it should explicitly issue a USE $PRINCIPAL:HUPENABLE. Also, ZSHOW "D" displays the HUPENABLE state for $PRINCIPAL. (GTM-DE222430)