source: trunk/txt/changes-log

- Testcase 11.4.2004 (latest MMOS/2 installation)
  Bug was in actual installation. Incorrect group was set on specified files.
  Anyway, correction got applied to mmi_basescr.c - setting GroupPtr for files
  specified as INI or config script in group-specification (control.scr) to
  actual group (original groupptr from files.scr-allocation is overwritten)
  This will fix the problem although the bug is in the installation-script.
- Fixed processing of cardinfo.dll and custom scripts. Now cardinfo.dll is
  processed AT FIRST and the custom scripts afterwards. That way cards that
  are generated by cardinfo and modified by custom scripts is possible.
  I'm not sure about original minstall behaviour though.
- Testcase 14.5.2004
  No error message was displayed, when GENIN.DLL got used but no cardinfo.dll
  was found. This was just confusing (no real bug). Error message is now shown.
- 5.6.2004
  Problem was in "REPLACEMENT" method for config-control-files. We didn't add
  ';' to the end of a path (if taken from DestPaths) and that is done by
  minstall. It seems that some applications require this weird behaviour.
- 13.6.2004
  GENIN.DLL has *undocumented* feature. ssdllinputparms may also contain a DLL
  name, that will get used instead of standard "cardinfo.dll". Previously we
  processed cardinfo.dll as hardcoded DLL.
- 15.6.2004
  genin compatibility code has a problem with install.dll from AudioPak
- 04.7.2004
  crash problem genin compatibility solved: IBM C Compiler bug. Inserted 2
   lines of code (MINSTALL_LogToFile) and it works now.
- 24.8.2004
  problem with timidity installation - Problem solved: bug was in installation
   script and original IBM installer worked because it's so buggy.
   Detailed information in bugs.timidity
- 5.12.2004
  problem with ELSAmotion installation - MINSTALL says "Bad value for PDDname"
   Problem solved: bug was in installation CARDINFO file. It specifies PDD name
    as ELSAMO1, but the limit is 6 chars for PDD names, because a sequential
    number and a '$' is appended afterwards. The limit is specified by IBM under
    MMPM/2 Device Driver Reference. Search for CARDINFO and look at Step 3:
    "Modifying the CARDINFO.RC file".
   Solution: Change ELSAMO1 to ELSAMO in CARDINFO.DLL
- 5.12.2004
  problem with Hauppauge Win/TV ISA - MINSTALL says "Illegal prompt count" on
    IBMOPTNS.DLL file
   Problem solved: DLL is using 24 possible answers on one prompt. Bug was in
    MINSTALL. Limit for prompts is 10, limit for answers is 25. I checked the
    answer-limit to be less or equal 10 instead of 25.
   Solution: Will change the code and recompile a version.

- 29.05.2005
  problem with ELSAmotion installation - MINSTALL says Bad Value for PDDname
   Problem solved: PDDname listed is 7 chars. 6 chars is the limit given by
   IBM, cause sequential number and '$' is appended. Listed in MMPM/2 Device
   Driver Reference. Anyway, I implemented code that will cut the PDDname to
   6 characters. This will also put a warning in MINSTALL.LOG.
   Solution: compatibility fix done.

  problem with ELSA PS2VideoIn - MINSTALL says Numeric value expected (28)
   Problem solved: MINSTALL did not accept numeric values, if a ')' was
    immediately behind the value.
   Solution: Bugfix done.

  further problem with ELSA PS2VideoIn - MINSTALL says DrvVSDTable missing.
   Problem solved: DrvVSDTable really missing in ini.ch - it was even commented
    out. Anyway, I have now switched some parameters to "optional" state. Still
    a warning is written into MINSTALL.LOG about that missing parameter. Also
    only a handful parameters are now optional.
   Solution: Compatibility fix done.

  problem with Xine - MINSTALL says DrvVSDTable/DrvPDDName missing.
   Problem solved: See ELSA PS2VideoIn for solution
   Solution: Compatibility fix done.

  problem with ESS Solo PCI driver - MINSTALL says Bad Value for ProductVersion
   Problem solved: Driver version is 6 chars wide. IBM limit is 5 chars.
   Anyway, I implemented code that will cut the productversion to 5 characters.
   A warning will get written in MINSTALL.LOG.
   Solution: Compatibility fix done.

  problem with Opti MAD 16 - MINSTALL hangs using 99% system usage
   Problem solved: MINSTALL does NOT hang, but DIALOG.DLL (opti supplied)
                    ACTUALLY GENIN.DLL gets called by DIALOG.DLL and that one
                    hangs, BECAUSE SOMEONE FORGOT TO INCLUDE THE FREAKIN
                    REPLACEMENT DLL INTO THE PUBLIC RELEASE AND PUT THE
                    ORIGINAL IBM CODE THERE - THIS COULD CAUSE SERIOUS HAVOC!!!
                    It is already -which means some packages won't be 100% CID
                    installable because of a popup by GENIN- All GENIN bugs may
                    occur here as well!!! Please also drop GENINMRI.DLL
                    completely from distribution and use my replacement (stub)
                    GENIN.DLL!! That one isn't doing anything at all and that's
                    how it's supposed to work. Argh, this one is really silly.
   Solution: Replacement GENIN.DLL has to be put into MMOS2\DLL directory. I
              also implemented code into MINSTALL that will detect my own GENIN
              and if it's not found, MINSTALL will refuse to work.

  Known problem solved: Exception that occured sometimes during Custom-API
                         phase was MCI-API/Stack related. Increased stack size
                         and exception vanished.

  Further improvements: There is also a new switch for MINSTALL/CLI:
                         /* - this one will force it to stay in CLI mode and
                               not give control to MINSTPM.EXE even if
                               possible. So debugging things is much easier.
                               MINSTPM.EXE also has problems when started on
                               normal packages and replies with an error
                               message.

                        Displayed return-code now also contains an human
                         readable description (most people don't know what
                         FE00 means in CID speech ;)

                        MINSTALL.MSG got extended. Please use new version
                         otherwise empty messages will appear as warnings.

- 26.06.2005
  problem with Opti MAD16 driver - custom DLL says "could not load DLL"
   Problem solved: Depends on current directory. Custom DLL loads it from
                    current directory. If /F is used, current directory is not
                    source directory and DLL loading will fail.
   Solution: Original MINSTALL needs to get checked, supposed changing to
              source directory of package when installing.
  problem with several drivers - hangs
   Problem solved: Probably - increased stack size of MINSTALL.EXE and problems
                    vanished, at least the ones that I was able to reproduce
   Solution: New MINSTALL.EXE

                        Latest version is: MINSTALL.EXE - v1.03
                                           MINSTALL.DLL - v1.03
                                           GENIN.DLL    - v1.01

- 04.07.2005
  problem with all sorts of packages - crash and hang after package
   installation
   Problem solved: MINSTALL_CleanUp() was closing log-file, even if none was
                    currently open. Logic behind CID installation needs 2
                    CleanUp() calls resulting in possible crash condition. Hang
                    occured because of IBM fclose() code.
- 05.07.2005
  problem with Aureal 8820 package - ASP4CTRL.DLL is not copied to \MMOS2\DLL
   Problem solved: New MINSTALL removes all custom DLLs that are mentioned in
                    CARDINFO. It only uses them for calling entrypoints during
                    installation. ASP4CTRL.DLL is therefore marked non-included
   Solution: Implemented white-list for DLL removal, that will stop DLL from
              getting blacklisted.
- 10.07.2005
  problem with Opti MAD16 package - crash when selecting Win-OS2 driver
   Problem solved: Crashes in WININST.DLL. It does some weird things to the
                    MINSTALL calling stack. Also the previous called DIALOG.DLL
                    (also package supplied) CLOSES ALL SORTS OF HANDLES, WHICH
                    MEANS PM IS NOT INITIALIZED ANYMORE.
   Solution: Put all custom DLL calls into another separate thread and
              call them from there. This should make everything work even if
              custom DLLs act like morons.
- 13.07.2005
  new features implemented:
   - Removes R/O destination file attributes, so overwriting works anytime
   - Sets master volume to 60 on base installation (only!), setting to 50
      doesn't work using MciSetSysValue()
   - Filters CONFIG.SYS entries added by custom DLLs. If they try to add bad
      files like TIMER0.SYS, CLOCK01.SYS, CLOCK02.SYS or RESOURCE.SYS. Action
      is logged.

- 14.09.2005
  problem with some packages, where scripts are not listed in file-listing
   Solution: Compatibility fix allows this to happen, if the script is in
              package root. In any case a warning is logged. Documentation of
              file-listing specifies: "...that lists all the installable files
              in the installation package." - so I still think that original
              MINSTALL accepting this case seems to me like another bad code
              issue ESPECIALLY because one may not specify path information on
              control scripts (e.g. ssinich=..\test.scr doesn't work, where in
              fact it would be only logical to handle this one, if one is not
              expected to list the script in file-listing.)
  reaction of minstall.exe on '/F:' parameter seems to be incorrect
   Case solved: behaviour works as designed. Original MINSTALL accepts ANYTHING
                 on commandline, never reports any problem. The /F switch is
                 supposed to specify a control.scr location. If one specifies
                 /F:d:\invaliddir\control222.scr original MINSTALL still doesn't
                 report any problem and gets control.scr from current directory
                 So I would say that new MINSTALL reacts in an absolutely
                 valid way. If one wants to use command-line version of
                 MINSTALL, one is supposed to use '/*' parameter (new feature
                 of new MINSTALL, because original MINSTALL didn't have CLI).
                 Specifying no parameter is SUPPOSED to work as well and read
                 control.scr from current directory, BUT doing so spawns the
                 GUI version (minstpm.exe). If that doesn't do its job
                 according what it's supposed to do, it's not my fault.
                 (It's supposed to use MINSTALL.DLL-API to read out package in
                  current directory and show that information to the user).
- 10.02.2006
  reported ssselect="VERSION" doesn't work as specified in IBM documentation
   Case solved: behaviour fixed, complist.ini writing was never done - it's now
                 implemented as well, but SUBSYSTEM_GROUPSIZE key is not
                 because there is no documentation about it anywhere and also
                 I can't fiddle out what is saved. As it's not documented, it's
                 probably only used by original MINSTALL internally and so it
                 shouldn't hurt not to write it.
                 Anyway, I save a value called "INSTALLED_BY"=="MINSTALL/REMAKE"
                  so one can distinguish who installed which package, if still
                  original minstall is also used. Note that this is of course
                  quite useless in eCS 1.2+.
  feature request: check build-level of files and never overwrite older ones
   rejected: some files do not have build-level information attached, most IBM
              ones contain "xx.yy" as version number, still this may be any sort
              of string in other files, so there would be no generic way of
              checking. Probably time-stamp comparing is more appropriate.