Re: [Freedos-kernel] Kernel patch policy, branches, and FreeCOM

2004-07-19 Thread Arkady V.Belousov
Hi!

19-Июл-2004 23:59 [EMAIL PROTECTED] (tom ehlert) wrote to Eric Auer
<[EMAIL PROTECTED]>:

>> I am sure 2035a fixes several bugs of 2035
te> found the ret_ah bug. if that has real world implications is unknown.

 This relates to disks formatting.

[...]
te> possibly made the graphical MENU better - but certainly broke Eric's
te> and mine non-graphical config.sys

 "Broke"?! Loud statement. All works.

[...]
te> IMO that's less then impressive

 You miss also:

- fixed compilation by MSVC.
- fixed (int inthndlr.c) some INT21 functions.
- fixed (in dsk.c) some IOCTL functions.
- fixed VERSION=.
- fixed config.sys processing (for example, 0x1A character was treated as
  space).
- PSP now generated (INT21/26, INT21/55) same, as in MS-DOS.

This is not counting a lot of other small fixes, improvents (for example,
config.sys processing now is much stronger and more mistypes now diagnosed)
and optimizations (Lucho reports, that his image now less than 40k; in
memory kernel now uses also less memory, especially each environment is
shorter).




---
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_idG21&alloc_id040&op=click
___
Freedos-kernel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/freedos-kernel


Re: [Freedos-kernel] Kernel patch policy, branches, and FreeCOM

2004-07-19 Thread tom ehlert
Hello Eric,

> I am sure 2035a fixes several bugs of 2035

found the ret_ah bug. if that has real world implications is unknown.

found the Ulong 32%32 bug - impressive, but 100% irrelevant to the
kernel as all real world use will work 100% ok.

claims to have found a bug in umb_link code for a001:
  now it doesn't work either, and is again irrelevant as noone uses
  (for good reasons) /I=a000-afff

possibly made the graphical MENU better - but certainly broke Eric's
and mine non-graphical config.sys

found a bug so now MS Smartdrv loads - unfortunately now FreeCOM
doesn't work any longer (even if that's a bug)

IMO that's less then impressive


just as a OT sidenote, found somewhere on joelonsoftware.com:
the old WIN developers took it *personally* if some old stuff wouldn't
run any longer. this went as far as:
  they found that SIMCITY would run beutiful on DOS, but crash in
  windows. after some lengthy debugging they found that simcity frees
  some memory - and uses it afterwards.
  sure this works in DOS - and fails badly in a multitasking
  environment where this memory is reused by someone else.

now there 2 solutions for that

  solution a) say this code is buggy, ask manufacturer

  solution b) implement 'is the current task simcity ?'
  if so, delay the *real* free for a couple of seconds.

they (the MS developers) took b)

and for me this ends the threat.

tom

















---
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_id=4721&alloc_id=10040&op=click
___
Freedos-kernel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/freedos-kernel


[Freedos-kernel] Kernel patch policy, branches, and FreeCOM

2004-07-19 Thread Eric Auer

Hi all, Arkady:

PLEASE do make FreeDOS 2035a behave as "wrong" as 2035, because
people will want to compare 2035a to 2035, and if things happen like

- you cannot format unformatted drives (disk access flag broken)
- stable version of FreeCOM (0.82pl3, as opposed to 0.82pl3xyz testing)
  crashes if you try to use F5 at startup

then it is very hard to get the motivation to do experiments with 2035a.
If you insist on cloning MS behaviour for F5: You can add a SYS CONFIG
variable temporarily, which has to default to FREEDOS behaviour. Then
those people which NEED the MS behaviour (do they exist? Why is it needed?)
can still enable that mode, whil other people will simply be able to use
2035a as a replacement for 2035 and check if things get better by that.

I am sure 2035a fixes several bugs of 2035, but please tell us in great
detail which of your many patches actually fix bugs, and which! Most of your
patches contain improved comments and small optimizations. If one such patch
also fixes a BUG, it is usually very hard to notice, especially if your
patch is very LONG but your comment is very SHORT. Your patch announcement
mails often contain at most 5 lines of explanation each, if at all.

Another thing which would be really GOOD to have: Please let Lucho put all
patches online, either separately or in one ZIP/RAR/TGZ/..., along with a
SEPARATE text file which tells for every single patch file
- when you sent it to freedos-kernel / what the file date is
- whether and which bugs it fixes
- whether it contains comment improvements
- whether it contains optimizations and whether those have been tested to
  be 100% compatible with the old implementation

Thanks a lot! That should allow Jeremy and Tom to use more of your patches
for the STABLE branch of the kernel, and it should allow people who are
interested in the DEVEL / cutting edge branch of the kernel to quickly find
the corresponding patch if they find a bug. Example: I found that the changed
dsk.c or something related to it no longer honours the "disk access flag", so
it is impossible to format disks which do not already have a valid boot sector
now. Because it is too hard to find out which patches affected dsk.c in what
way, I would probably simply replace the WHOLE dsk.c - or even the whole kernel
if I would have been able to pinpoint the problem to "probably dsk.c" - by the
classic 2035 version again, just to get FORMAT working again.
If, however, it is documented which patch does what, and if the patches are
somehow available for easier download than digging deep in my mail archives,
then I can REVERSE APPLY single patches until I find the one patch which broke
format for me. This will also make it easier for you and me to repair things,
and will indirectly improve acceptance of your patches for the STABLE kernel.

Eric



---
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_id=4721&alloc_id=10040&op=click
___
Freedos-kernel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/freedos-kernel