Hi, still having same SM bug with 0.82.23 🙁 at least on my system: Debian x64 and dosbox-x compiled with ./build-debug

sm segment:offset af -> byte becomes 00
sm segment:offset bf -> byte becomes BF (OK)
sm segment:offset cf -> byte becomes 00
sm segment:offset df -> byte becomes 00
sm segment:offset ef -> byte becomes EF (OK)
sm segment:offset ff -> byte becomes FF (OK)

Same behaviour with "sm segment offset" syntax.

If you need more details or something you can ask, or if I can do something for fixing it.

Regards.

It may not be reproducible, but where a large time delay occurs at startup (such as loading cd audio), then sometimes movement of the parent window before the shell loads will lead to a sudden exit.

Well, I've manage to locate where the problem is.

For anyone interested the following patch will let you write bytes AC, AF, CF and DF in memory.

1--- debug.cpp.ORIG	2019-12-01 01:27:21.000000000 +0100
2+++ debug.cpp	2019-12-04 13:19:02.243257342 +0100
3@@ -1330,10 +1330,6 @@
4     else if (strncmp(hex, "FS", 2) == 0) { hex += 2; regval = SegValue(fs); }
5     else if (strncmp(hex, "GS", 2) == 0) { hex += 2; regval = SegValue(gs); }
6     else if (strncmp(hex, "SS", 2) == 0) { hex += 2; regval = SegValue(ss); }
7-    else if (strncmp(hex, "AC", 2) == 0) { hex += 2; regval = GETFLAG(AC); }
8-    else if (strncmp(hex, "AF", 2) == 0) { hex += 2; regval = GETFLAG(AF); }
9-    else if (strncmp(hex, "CF", 2) == 0) { hex += 2; regval = GETFLAG(CF); }
10-    else if (strncmp(hex, "DF", 2) == 0) { hex += 2; regval = GETFLAG(DF); }
11     else if (strncmp(hex, "ID", 2) == 0) { hex += 2; regval = GETFLAG(ID); }
12     else if (strncmp(hex, "IF", 2) == 0) { hex += 2; regval = GETFLAG(IF); }
13     else if (strncmp(hex, "NT", 2) == 0) { hex += 2; regval = GETFLAG(NT); }

When I type "SM some:addr XX" I want to write opcode XX in memory, not 0 or 1 based on some flag value for XX=AC, AF, CF or DF.

Patch tested with last dosbox-x release: 0.82.24

Happy debugging 😀

I believe the register/flag name parsing was originally intended to be for SR only, not for SM.

Does it work to write register values into memory (e.g. SM 0000:0200 AX)? If so, it may have been intended as a feature, but the collision between flag names and hex values could have been overlooked.

Yes, you can write register values into memory.

Found a bug in INT 10h AX=101Ah that was fixed for a game.

The function in DOSBox behind it reads the Attribute Controller in a way that leaves the screen blank.

You'd never notice in DOSBox SVN because SVN does not correctly emulate the display blanking caused by attribute controller programming (bit 5).

Prior to the change the function INT10_GetDACPage() was exactly the same in DOSBox-X as it was for DOSBox SVN.

https://github.com/joncampbell123/dosbox-x/issues/1362

I actually saw some of the issues on the projects github page, and one of them was a problem to run Photoshop on Win 3.x. Well, I have the same issue. It works for me with DosBox and cputype pentium_slow, but only after 2 or 3 attempts to run the software. Meanwhile, in DosPad, a port of dosbox for iOS, I can’t make it work. Since dospad is a very old port, is full of errors. But at least it runs my games and win 3.x.

I believe the issue (not only for dospad, but also for any version of dosbox including yours) is related to the cpu type.

While on vmware fusion, I can run the software well.

Edit: for what it is worth, I was attempting to run versions 3.0.4 and 3.0.5. All give me the same problem with win32s (even though I had it installed).