The genRecieve code is UGLY, primarily because I hacked in as an
afterthought, The compiler was not initially designed to pass parameters
As for 2.2.1 there is really nothing holding us. I will make a tar.gz of the
source today. Dave I forgot where you want the file so that you can put
it on the webpage.
I need help building the Windows binaries. I don't have cygwin installed
on any machine.....
I will also tag CVS with the version number..... The tag will be V_2_2_1.
[mailto:sdcc-devel-admin@... Behalf Of Kevin Vigor
Sent: Thursday, October 12, 2000 10:23 PM
Subject: RE: [sdcc-devel] Post 2.2.1 considerations. Was: Serious bug
> So, a construction like:
> mov dptr,#_ReadPCF8591_address_1_1
> mov a,dpl
> movx @dptr,a
> obviously shows a flaw that still exists in the code-generator and
> not in
> the parser/optimizer. No offence meant.
Oh yes, there's clearly a bug; we shouldn't be doing the genReceive
case at all when all accesses to the parameter have been optimized
away. I hadn't noticed just how screwed-up that genReceive code
was... no offense taken, that's some ugly stuff!
I'm only resisting fixing this because the workaround is easy once
you know the problem: don't have any meaningless references to a
function parameter (if there are no references at all, it's fine; if
the parameter is actually used, it's fine; it's only the case in
which a function parameter is referenced in a useless way that the
optimizer can remove that we have a problem, and that case can easily
be avoided). So I don't think this bug is very serious. But it is
definitely a bug, and I will do my best to fix it.
(by the way, you don't want to know how long it too me to spot that
semi-colon! I learned Pascal first (showing my age...) and my
brain still just completely ignores semi-colons when reading...)
> am waiting for the 2.2.1 cut as well before ci-ing it. So, please
> let's get
> on with it.
I'd agree; Sandeep, anything we can do to expedite cutting a 2.2.1
> Now that I know that there IS a "known bugs" listing (there is even
> a "task"
> list!) I wonder why we don't really use it.
Just not in tha habit, I guess. One thing: please send mail to the
-devel list if you create a new bug/task (there were some bugs in the
ds390 stuff that had been reported for months, and I had no idea that
the bug reports were there; I wish Sourceforge had an option to mail
me when a bug was created, but if there is one, I surely can't find
sdcc-devel mailing list