-This is the VMS version of GNU Make, updated by Hartmut Becker
-
-Changes are based on GNU make 3.82.
-
-This version was built and tested on OpenVMS V7.3 (VAX), V7.3-2 (Alpha) and
-V8.3-1H1 (I64).
+This version of GNU make has been tested on
+OpenVMS V8.3 (Alpha) and V8.4 (Integrity).
Build instructions
+------------------
Make a 1st version
$ @makefile.com ! ignore any compiler and/or linker warning
- $ rena make.exe 1st-make.exe
-Use the 1st version to generate a 2nd version
- $ mc sys$disk:[]1st-make clean
+ $ copy make.exe 1st-make.exe
+ Use the 1st version to generate a 2nd version
+ $ mc sys$disk:[]1st-make clean ! ignore any file not found messages
$ mc sys$disk:[]1st-make
-Verify your 2nd version
- $ rena make.exe 2nd-make.exe
+ Verify your 2nd version
+ $ copy make.exe 2nd-make.exe
$ mc sys$disk:[]2nd-make clean
$ mc sys$disk:[]2nd-make
+ Don't use the HP C V7.2-001 compiler, which has an incompatible change
+ how __STDC__ is defined. This results at least in compile time warnings.
+\f
+Changes since GNU make 3.82
+---------------------------
-Changes (3.81.90)
+Fix build problems.
+
+The new feature "Loadable objects" is not yet supported. If you need it,
+please send a change request or submit a bug report.
+
+The new option --output-sync (-O) is accepted but has no effect: GNU make
+for VMS does not support running multiple commands simultaneously.
+\f
+Changes for GNU make 3.82
Michael Gehre (at VISTEC-SEMI dot COM) supplied a fix for a problem with
timestamps of object modules in OLBs. The timestamps were not correctly
Build fixes for const-ified code in VMS specific sources.
-Build notes:
-- Try to avoid HP C V7.2-001, which has an incompatible change
-how __STDC__ is defined. This results at least in compile time warnings.
-
-- On V8.3-1H1, if you press Ctrl+C you may see a traceback, starting with
-%SYSTEM-F-CONTROLC, operation completed under CTRL/C
-%TRACE-F-TRACEBACK, symbolic stack dump follows
-image module routine line rel PC abs PC
-
-DECC$SHR C$SIGNAL gsignal 27991 0000000000001180
-FFFFFFFF84AB2DA0
-DECC$SHR C$SIGNAL raise 28048 0000000000001280
-FFFFFFFF84AB2EA0
-DECC$SHR C$SIGPENDING decc$$deliver_signals
- 12475 0000000000000890
-FFFFFFFF84C13690
-...
-This looks like an incompatibility to the Alpha and VAX behavior, so it looks
-like a problem in I64 VMS version(s).
-
-- There is no clean build on VAX. In the environment I tested, I had to use GNU
-make's alloca which produced a couple of compile time warnings. It seems too
-much effort to work on a clean build on VAX.
-
A note on appending the redirected output. With this change, a simple mechanism
is implemented to make ">>" work in action lines. In VMS there is no simple
feature like ">>" to have DCL command or program output redirected and appended
associated command procedure is left in SYS$SCRATCH as well. Its name is
CMDxxxxx.COM.
-Change in the Ctrl+Y handling
-
-Ctrl+Y was: The CtrlY handler called $forcex for the current child.
-
-Ctrl+Y changed: The CtrlY handler uses $delprc to delete all children. This way
-also actions with DCL commands will be stopped. As before Ctrl+Y then sends
-SIGQUIT to itself, which is handled in common code.
+Change in the Ctrl+Y handling. The CtrlY handler now uses $delprc to delete all
+children. This way also actions with DCL commands will be stopped. As before
+the CtrlY handler then sends SIGQUIT to itself, which is handled in common
+code.
-Change in deleteing temporary command files
-
-Temporary command files were deleted in the main line, after returning from the
-vms child termination handler. If Ctrl+C was pressed, the handler is called but
-did not return to main line.
-
-Now, temporary command files are deleted in the vms child termination
-handler. That deletes the them even if a Ctrl+C was pressed.
+Change in deleteing temporary command files. Temporary command files are now
+deleted in the vms child termination handler. That deletes them even if
+a Ctrl+C was pressed.
The behavior of pressing Ctrl+C is not changed. It still has only an effect,
after the current action is terminated. If that doesn't happen or takes too
long, Ctrl+Y should be used instead.
\f
-Changes (3.80)
+Changes for GNU make 3.80
. In default.c define variable ARCH as IA64 for VMS on Itanium systems.
. In makefile.vms avoid name collision for glob and globfree.
-
-In newer version of the VMS CRTL there are glob and globfree implemented.
-Compiling and linking may result in
-
- %ILINK-W-MULDEFLNKG, symbol DECC$GLOBFREE has subsequent linkage definition
- in module DECC$SHR file SYS$COMMON:[SYSLIB]DECC$SHR.EXE;1
- %ILINK-W-MULDEF, symbol DECC$GLOBFREE multiply defined
- in module DECC$SHR file SYS$COMMON:[SYSLIB]DECC$SHR.EXE;1
-
-linker messages (and similar for DECC$GLOB). The messages just say, that
-globfree is a known CRTL whose name was mapped by the compiler to
-DECC$GLOBFREE. This is done in glob.c as well, so this name is defined
-twice. One possible solution is to use the VMS versions of glob and
-globfree. However, then the build environment needs to figure out if
-there is a new CRTL supporting these or not. This adds complexity. Even
-more, these functions return VMS file specifications, which is not
-expected by the other make sources. There is a switch at run time (a VMS
-logical DECC$GLOB_UNIX_STYLE), which can be set to get Unix style
-names. This may conflict with other software. The recommended solution
-for this is to set this switch just prior to calling main: in an
-initialization routine. This adds more complexity and more VMS specific
-code. It is easier to tell the compiler NOT to map the routine names
-with a simple change in makefile.vms.
-
-Some notes on case sensitive names in rules and on the disk. In the VMS
-template for CONFIG.H case sensitive rules can be enabled with defining
-WANT_CASE_SENSITIVE_TARGETS. For recent version of VMS there is a case
-sensitive file system: ODS5. To make use of that, additionally un-defining
-the HAVE_CASE_INSENSITIVE_FS is required. As these are C macros, different
-versions of make need to be built to have any case sensitivity for VMS
-working. Unfortunately, for ODS5 disks that's not all.
-
-- Usually DCL upcases command line tokens (except strings) and usually the
- file system is case blind (similar to how Windows systems work)
- $ set proc/parse=extended/case=sensitive
- preserves lower and UPPER on the command line and (for this process and all
- sub-processes) enables case sensitivity in the file system
-
-- Usually the CRTL tries to reverse what DCL did with command line tokens, it
- lowercases all tokens (except strings)
- $ define DECC$ARGV_PARSE_STYLE enable
- passes (the now preserved) lower and UPPER from the command line to main()
-
-- Usually the CRTL upcases the arguments to open() and friends
- $ define DECC$EFS_CASE_PRESERVE enable
- preserves the names as is.
-
-It is important to know that not all VMS tools are ready for case sensitivity.
-With this setup some tools may not work as expected. The setup should not
-blindly be applied for all users in default login procedures.
-
-Example? The poor coding gives a compiler message, showing that there are
-different files:
-
-$ dir
-
-Directory ODS5DISK[HB]
-
-A.c;1 B.c;1 c.c;1 X.c;1
-x.c;1
-
-Total of 5 files.
-$ ods5make x.obj
-cc /obj=x.obj x.c
-
-foo(){lowercase_x();}
-......^
-%CC-I-IMPLICITFUNC, In this statement, the identifier "lowercase_x" is implicitly declared as a function.
-at line number 1 in file ODS5DISK[HB]x.c;1
-$ mc SYS$SYSDEVICE:[HARTMUT.MAKE_3_80P]ods5make X.obj
-cc /obj=X.obj X.c
-
-foo() {UPPERCASE_X();}
-.......^
-%CC-I-IMPLICITFUNC, In this statement, the identifier "UPPERCASE_X" is implicitly declared as a function.
-at line number 1 in file ODS5DISK[HB]X.c;1
-$ dir
-
-Directory ODS5DISK[HB]
-
-A.c;1 B.c;1 c.c;1 X.c;1
-x.c;1 X.obj;1 x.obj;1
-
-Total of 7 files.
-$
\f
This is the VMS port of GNU Make done by Hartmut.Becker@compaq.com.
projects / thirdparty / make.git / commitdiff
