Z180DMA uzero causes regression: login fails with error message

[nkoch61]

I've built a diskimage for sc126 under Ubuntu 18.04.

When logging in as root I see:

Welcome to FUZIX.
: is not an identifier

This is an error message from sh.
My guess is that it comes from setname(),
called by sh_getenv().

Single user boot works.

The pre-built image does not show this error.

[nkoch61]

I've put some print statements into sh and init. It seems like init produces empty environment entries which sh does not accept.

[EtchedPixels]

I changed the Z180 code to use DMA for the user memory transfers. Does reverting commit 861ed42 fix it ?

[EtchedPixels]

Not sure what that is trying to say/show ?

We know that

1. init never adds more than 9 variables (not that a sanity check isn't a bad idea)
2. entries never get blanked or duplicated so the loops should have no effect
3. the array starts zeroed so setting an end NULL is not needed

If that does make a difference it would mostly suggest env[] isn't clear (which would be perhaps a bug in the DMA changes although I can't duplicate it here at the moment), or a compiler specific problem (either a bug in the code elsewhere the compiler happens to show up or a compiler bug).

[nkoch61]

My knowledge is not deep enough to understand in which contextes (forked, exec'd) this environment is used. It's just that after these changes my login problems went away.
I had modified init and sh to print their environments. And I saw empty strings at different places in the environment.
But I will follow your suggestion above and write about the results.

[EtchedPixels]

What sdcc version is in your Ubuntu (sdcc -v) ?

[nkoch61]

After lots of syntax errors with "standard" sdcc as installed via apt, I've installed https://github.com/EtchedPixels/sdcc280.
Version is 3.9.1.

[nkoch61]

I changed the Z180 code to use DMA for the user memory transfers. Does reverting commit 861ed42 fix it ?

Yes, it works.

[EtchedPixels]

Thanks. I will go back over that and try and reproduce it here.

[nkoch61]

I forgot to mention two things:

• My board is an SC130, not an SC126.
• I have an RTC board DS1302 installed, with address set to 0x10.