ADC setup code is strange for 16F877 (PCM 3.202)

[PCM programmer]

PCM vs. 3.202 is doing something weird with the ADC setup functions.
Instead of writing to ADCON0 and ADCON1 as a single byte operation,
the compiler is doing it one bit at a time. This must have something
to do with the major changes that CCS did for vs. 3.200.

What I dislike about this method is that it cycles the Port A pins
through several different modes before they reach the intended mode.
That's just wrong on a philosophical level. It could potentially
cause some problem with external circuits by switching a pin into
a mode that's not intended.

In addition to that, I noticed that the startup code for ADCON1 is
probably incorrect. Traditionally, CCS has initialized Port A as all
digitial. But look at the code right below main(). They're doing
a BSF on bits 0-3 of ADCON1. That sets it to 0x0F, which has AN0
as analog, and AN2 and AN3 as Vref pins. For years, they have
initialized ADCON1 to 0x07 (all digital). Why this change ?
I have to assume it's a mistake.

The code below was compiled for a 16F877 with PCM vs. 3.202:

[Guest]

It must be a mistake, indeed. The ADC with 3.202 not works properly.
With a well-tested source code I got 0 for all channels. Downgrading to 3.191 and everything is fine.

[PCM programmer]

[Guest]

I have tested this with:
16F877
18F452
12F675
getting the same result. It must be something "orbital" about the new ADC handling style.

[Haplo]

Thanks for sharing this, guys...Yesterday I was thinking about updating my compiler from 3.190 to 3.202 in a 16F877 based project with a lot of ADC code, but thanks to your warnings I'll hold off.
Had I changed it, I would've spent days trying to figure out what was wrong. A broken ADC code is the last thing you expect from the new version of the compiler, afterall.

[Kieran]

I have just recompiled a project for a 16F876 using ADC and it works fine.

[PCM programmer]