.
.
WARNING: this is obsolete, IDE from 1.67.35 has most the bugs fixed + INI support + directory browsing + debug support, just no [F1]
.
.

Updated DOS IDE : Download now : (160 KiB) (note: based on 1.67.29 package, not up to date)

.
.

Well, almost 2 years elapsed since Tomasz fixed last 2 bugs in FASM DOS IDE pointed by me, see here (2007-Jan), so the time has finally come to fix a bit more The PF bug from 2008-Mar got fixed by Tomasz (just 2 bytes changed) in 1.67.19 release, see here (2008-Nov), also the memcorrupt bug (discussed by PM only) got fixed (just 2 instruction added) by him and is included here. Many other things (bugs + optimizations + enhancements) got fixed by me, most of them never mentioned here before, most notably the [F1] thing I had missed so badly I didn't fix the "DEL-BUG" , see here (2008-Jan) because it's a feature, not bug .

Known FASM DOS IDE issues:

BUG (fixed) Memory corruption 18...140, fixed by Tomasz after 1.67.29, thus fix not included in 1.67.29, but in 1.67.32 and here it is

BUG (fixed) Page Fault bug in line 1010, see here (2008-Mar), fixed by Tomasz in 1.67.29

BUG (NOT fixed) S&R works badly, bug shared with Win32 IDE, see here (2007-Aug)

BUG (fixed) Always hanged on 8086 (no VGA jump is near, no DPMI jump is near -> POPE CS !!!) fixed in 1.67.32 also

BUG (fixed) Code handling DPMI selector brewing failure assumed "True DPMI" , thus it printed garbage with CWSDPMI

BUG (fixed) "_copyright" defined but not used anywhere

BUG (fixed) Ctl-N and Ctl-BckSpc working but undocumented also fixed in 1.67.32

COS (fixed) "draw_centered_window" centered badly

COS [fixed) "draw_window" title not nice enough

OPT (fixed) Cursor hiding was inefficient and messy

OPT (fixed) "number_as_text" was very messy, inefficient and bloated, fault shared with Win32 IDE and commandline versions, thus contributing 25 bytes saving into the war on 64 KiB limit in CS in unreal mode !!!

OPT (fixed) Time boasting code was more bloated than necessary, fault shared with Win32 IDE and commandline versions, thus contributing 7 more bytes saving into the war on 64 KiB limit in CS in unreal mode !!!

ENH (fixed) Text dump of EXE looked badly, see 2 lower shots, see also "No version" issue and [F1] issue below

ENH (fixed) No version, fixed, see also [F1] issue below

ENH (fixed) Help & About on [F1], see 2 upper shots

ENH (semi-fixed) Vertical selection: implemented by Tomasz from the beginning, but hidden, partially "fixed" with my [F1] fix, still missing hor/ver status in bottom info line

ENH (NOT fixed) Directory browsing, Win32 IDE has "external" support

ENH (NOT fixed) Mouse support, Win32 IDE has "external" support

OPT (NOT fixed) Segmented non-ZERO based memory model

ENH (NOT fixed) Requires a DPMI host

COS (fixed) Suboptimal "No DPMI host" message, pointed 1'000'000'000 times by this guy

ENH (NOT fixed) "predefinitions" not supported, nor in Win32 IDE

ENH (NOT fixed) Debug output not supported, nor in Win32 IDE

ENH (NOT fixed) Compiler options not supported, partially in Win32 IDE

ENH (NOT fixed) Usage of .INI file not supported, unlike Win32 IDE

ENH (NOT fixed) Progress indicator for load and save

Changes:

(- 9 bytes) Fixed "draw_centered_window" centering, in "draw_window" moved box title by 1 char to the right -> nicer, cursor hiding clean up: optimized "update_cursor" , renamed "cursor_out_of_sight" -> "hide_cursor", promoted the piece to a sub, ensured EDX and EDI preservation, deleted cursor hiding from "compilation_directory_ok" and "message_drawn"

(+ 4 bytes) Applied Tomasz's patch for the memcorrupt bug

(+ 32 bytes) Re-added test for 80386, optimized VGA test, optimized DPMI selector brewing, fixed "True-DPMI-bug" , nicer and more informative complaints, better "No DPMI" message, large rearrangements resulting in 99% of jumps short instead of 99% near, obfuscated real mode complaints, see 2 bottom shots why

(+ 51 bytes) Updated VERSION.INC (a few files like FORMATS.INC do need adjustments, this adds version into DOS IDE, also should finally fix the versionning chaos and expired copyright trouble) , optimized usage of VERSION.INC, optimized "startup_failed" jumps and some "mov ecx"-stuff, adjusted "draw_caption"

(+ 992 Bytes) Implemented Help & About (for [F1] key see below), also gets displayed at startup if no files loaded, see 2 upper shots

(- 15 Bytes) Added support for [F1] in key evaluation and optimized a few jumps, also moved "call ascii_table_window" stuff

(- 25 bytes) Optimized "number_as_text" and renamed it to "number_to_string_dec"

(- 7 bytes) Optimized time boasting code

(no bytes) Added some comments

(no bytes) Updated (C)

(dont count) Updated TFM file

--------------

(+ 1023 Bytes) total bloat effect of my changeset

95'449 Bytes final EXE file size

This is a result one can't avoid to be very proud of, considering the 1110 Bytes size of only the raw key index text I added !!! Of course I could have pushed it below a decimal drivemaker's "kilo Byte" of <= 1000 Bytes as well with a few more dumb changes, though I discarded them now in favor of smarter optimizations in future.

I raised the compiler core version to 1.67.30 and IDE version to 0.93.16 , what I would prefer to see in Tomasz's next release is 1.68 and 0.94 of course

.
.



My new key index

.
.



My new "About" box

.
.



Any other development tool (compiler + editor) can compile itself within 0.2 seconds ??? It was at a Pentium ... would be even cooler to test on an 80386 running at 10 MHz ... I'm confident that FASM would pass

.
.



When looking into "old" FASMD.EXE from 1.67.29, one sees only what can go wrong

.
.



When looking into my new improved executable, one sees what it is (supposed to be) and what version

.
.

(Offtopic) Known and unfixed FASM DOS commandline version issues:

- Unreal mode and 64 KiB limit in CS (as of 1.67.28 from 2008-Nov the
reserve was cca 256 Bytes with listing hack, 1.5 KiB without)

- Gets no memory and crashes if RM, DPMI host present, no XMS, see here (2008-Feb)

- "number_as_text" and time boasting code inefficient (see IDE issues above)

- Debug output doesn't default to a useful filename

I tested latest 1.67.35 on my old 486 Sx/25 w/ 8 MB total RAM running MS-DOS 6.22 (instead of confusing the issue with any number of FreeDOS kernels). The fastest I seemed to be able to reassemble FASM.EXE itself was 9.7 secs (running on RAM drive without cache turned on). Normally it took about 11 or 13 secs on HD itself. The slowest was when I turned off the "turbo" button: 38.5 secs (which makes no sense, what speed is the cpu running at then??? I thought 16 Mhz, but apparently not). Just FYI.

EDIT: Oops, of course I meant 1.67.35.

No IDE ? Obviously with IDE it takes longer



??? Any benefit of caching RAMDISK to expect ?



Why ?



RTFM or test

Okay, so here's some more results in case anybody is interested:



That's basically XMS (HIMEM.SYS) vs. XMS+EMS (HIMEM + EMM386) on my 486 Sx/25 w/ 8 MB RAM (total) running from the HD. The second two are the same but with a RAM drive (SHSURDRV from FDRV-3.ZIP).

EDIT: The date is wrong (2-27) because the CMOS is dead, so ignore that, I don't have a time machine.

COOL. Takes cca 10 s on 80486. Having also the IDE compiled and with a free kernel would be even better ... no EMM386, just HDPMI32 (from IDE) vs Unreal (from commandline).



Better dead CMOS than dead OS

Okay, not to make a big deal out of it, but it seems to me that 1.67.37 lacks the flat real mode hack again. Is it not still planned for 1.68? Oh well, no huge loss I guess. It's a cool hack even if we don't really really need it (barely faster).

That's only temporary. I assure you 1.68 is going to have unreal working.

Then I wonder why you didn't apply my 2 (see above)



patches yet