SjASMPlus от z00m

[zebest]

Ped7g, снова требуется немного разъяснить

16+

Три ошибки:

и еще одна ошибка:


я так понимаю next-команды процессора в основном работают??
Тогда что может быть не так с этими четырьмя??
Это не на NEXT, но процессор он него.
Или "и так сойдет..." ?

[свернуть]

[Advertiser]

[NEO SPECTRUMAN]

Ped7g, а можно сделать такую команду
на подобии define но с безусловной заменой
и чтоб можно было менять даже пробелы

а то мощности define не хватает


например можно будет делать так

	if aa = 0
	replace "AAA BBB", "10 LD a"
	replace "VERSION", "0001"
	endif

	if aa = 1
	replace "AAA BBB", "10 LD (HL)"
	replace "label", "lb"
	replace "VERSION", "0002"
	endif

	if aa = 2
	replace "labelAAA", "label"
	replace "AAA BBB", "10 LD (HL)"
	replace "VERSION", "0003"
	endif



labelAAA BBB,B

	SAVESNA "000-VERSION.sna", $8000

aa = 0

label10 LD a,B

	SAVESNA "000-0001.sna", $8000

aa = 1

lb10 LD (HL),B

	SAVESNA "000-0002.sna", $8000

aa = 2

label BBB,B

	SAVESNA "000-0003.sna", $8000

[Ped7g]

https://github.com/z00m128/sjasmplus...es/tag/v1.15.0

v1.15.0 changelog:
* added `BPLIST` and `SETBP` to export breakpoints info from asm for Unreal and ZEsarUX emulators
* added ZX-like devices with 2/4/8 MiB of virtual memory
* the fake-sysvars/state of ZXSPECTRUM48/128/... devices reworked, moving stack down by default
* behaviour of `--fullpath` option unified across all platforms and compilers
* `DEFARRAY` has new operator `[#]` to retrieve current size of array.
* `MMU` has new optional third argument to set also address (like `ORG`)
* use of forward reference in `IF`/`IFN` emits only warning, and can be suppressed
* internal Lua updated to 5.1.5 (last official 5.1 version)
* STDIN can be read multiple times (per each "-" argument)
* new macro/lua examples: `sj_sysvars.i.asm`, `section.asm`, `lua_sin_table.asm`, `union_like_structures.asm`
* RAM limit exceeded warning/error reworked to report with more sense, fixed bug with missing labels
* `SAVETRD` warning about invalid extension can be suppressed by "; ok"
* `EMPTYTRD` takes as second argument disc label
* added `--outprefix` option to prefix any output-directive filename (issue #102)

Enjoy.

[NEO SPECTRUMAN]

* internal Lua updated to 5.1.5 (last official 5.1 version)

ВНЕЗАПНО adp полностью пережил переход на новую версию
готовый trd полностью совпал

название диска тоже работает

[Ped7g]

снова требуется немного разъяснить

If you click the letter of the particular test, it will show some debug values (usually the values which failed the test in the code).

With that info, check the test source code to figure out what precisely failed.

These four seems to be all I/O related, so it may be also the test does use some unlucky choice for port number to test against.

I don't remember the implementation details for these tests, it's long time since I wrote it.
If you can provide me with the debug info values of failed test, I may try to quickly look at the source what it means.

But yes, your Z80N seems to be mostly working correctly, if only these four are failing.

[zebest]

ну да, в этих кодах ошибок уж точно я ничего не пойму

16+

[свернуть]

Ну и регистры на всякий случай картинку.
Я думаю, что не хватает какого то жизненно важного регистра. Все то они мне не нужны.
Вот только с каким номером именно надо мне регистр))

[Ped7g]

в этих кодах ошибок уж точно я ничего не пойму

nextreg *r, *n:

It does test four NextRegister by writing/reading the value: LAYER2_XOFFSET_NR_16, VIDEO_INTERUPT_VALUE_NR_23, PALETTE_INDEX_NR_40, SPRITE_TRANSPARENCY_I_NR_4B
(it's enough just to store and then return the same value for the test, it's not testing if the next register actual functionality is implemented)

And your machine does also modify value written to I/O port 243B (value TRANSPARENCY_FALLBACK_COL_NR_4A is written there) - it should be not modified, "nextreg" instructions are independent from what is done directly to the $243B and $253B I/O ports.

So to simplify when the test did stop:
out ($243B),$4A
nextreg $16,$01
assert $4A == in ($243B) ; fails (is different)
assert $01 == nextreg_16 ; fails (is $00)

I guess the nextreg *r, A is the same issue.

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

OUTINB:

I/O port $243B is used for the test
out ($243B),$00
assert $00 == in ($243B) ; fails, reads $FF

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

JP (C):

Again depends on correct functionality of $243B / $253B, jumps to wrong address if port returns $FF instead of $00.

[Ped7g]

This doesn't fit well with current sjasmplus code-base ... I guess the least problematic way would be to add extra search+replace step before substitutions (DEFINE) are parsed, so it's possible, but it will slow down assembling of everything and it will be completely new code ( = new bugs ).

But I'm also not sure if this is good addition to assembler, I would personally probably use `sed` as first step ahead of assembling (to create temporary asm file), as the sed can do all these string replacements in much better way, including regex syntax to precisely target certain constructs (I did use `sed` while fixing one MSX project wanting to migrate from tniassembler to sjasmplus, to fix many small details about syntax of their sources, you can find the article on sjasmplus github wiki).

I would use something like (Makefile):

.PHONY all

all : aa0.tap aa1.tap aa2.tap

aa0.tap : main.asm
        sed -f sed_aa0_rules.txt main.asm > aa0.asm
        sjasmplus -Daa=0 aa0.asm

aa1.tap : main.asm
        sed -f sed_aa1_rules.txt main.asm > aa1.asm
        sjasmplus -Daa=1 aa1.asm

aa2.tap : main.asm
        sed -f sed_aa2_rules.txt main.asm > aa2.asm
        sjasmplus -Daa=2 aa2.asm

... building all variants with one `make` command.

... or something like that... there are also options to use `m4`, `awk` or even `cpp` (C pre-processor, not C++), depending on what kind of syntax/features you are looking for, all of these can be applied to asm sources easily.

Whatever hack-ish solution I will add to sjasmplus, it will be nowhere near the power of these specialized tools (unless I re-implement whole tool inside sjasmplus), so I think I will rather leave these non-trivial cases to the specialized tools (also with 30+ year history, so almost everyone knows them and most of their bugs were already found + fixed... contrary to sjasmplus).

[NEO SPECTRUMAN]

I would use something like (Makefile):

все таки больше бы хотелось именно find&replace
средствами самого ассемблера

[Ped7g]

The SAVETRD refactoring and "&" for sector-append of files is now at master branch in github.

If you are using SAVETRD in your projects, testing them now would be nice to catch any regressions before the v1.15.1 will be released.
(for example how to use the "&" check "tests/misc/trd/savetrd4.asm" or "docs/documentation.xml" (html is not updated yet)).

(of course you must build sjasmplus from the latest sources to have the new SAVETRD implementation)

- - - Updated - - -

все таки больше бы хотелось именно find&replace
средствами самого ассемблера

I don't understand what's the point of either adding lame sub-par limited feature in days/weeks of work (and that's still too much time), or trying to compete with excellent SW which already exists (sed/m4/awk/... all of that available everywhere, with years of expertise and know-how baked into them) - that would eat probably months/years of my time, completely wasted on something what I can run in 10 seconds.

So far I have seen no factual justification for such demand (and I would be very surprised if you would come with something meaningful, because I can't even imagine anything reasonable).