Re: ALR unable to get many packages

[Simon Wright <simon@pushface.org>]

Kenneth Wolcott <kennethwolcott@gmail.com> writes:

[About 70 lines of quoted comment. It'd help if you could leave in only
the relevant parts of the message you're replying to.]

> BTW, Simon, I am NOW using your ALR package. I am using the following
> build

I'm going to have to consider making releases in a way that makes it
easier for me to identify which release you're using.

> alr --no-color build -- -XTARGET=arm64-apple-darwin22.6.0

If this is when building aws, where there is an issue with TARGET, I'd
recommend setting TARGET to what you get by executing "gcc -dumpmachine",
which I believe should be "aarch64-apple-darwin21". The problem is that
aws's alire.toml sets TARGET to a value different from that needed by
the actual code.

>  I was able to "alr get" and successfully build only eight of the alr
> packages that I was interested in.

Sorry, what I'm interested in at the meoment is the ones that *didn't*
work because they tried to use an external that isn't set up yet;
something like

   $ alr with libreadline
   ⓘ Synchronizing workspace...
   Nothing to update.          

   Requested changes:                                                     

      ✓ libreadline * (add)

   Changes to dependency solution:

      New solution is incomplete.
      +🔎 libreadline * (new,external)            <============

   Do you want to proceed?
   [Y] Yes  [N] No  (default is No)

> It seems clear that MacPorts and homebrew are not compatible
> (especially if building packages using either or both of them).

I think you could probably install both for getting tools (eg graphviz),
but to install libraries for building software against you have to
decide which one to go for. My latest version of alr (not yet released,
issue with Github CI) checks whether "brew" or "port" are on the PATH;
if it finds both, it chooses Homebrew.

> BTW: mathpaqs failed to build due to "digits value out of range,
> maximum is 15".

This is because GNAT on x86_64 implements Long_Long_Float as 128 bits,
'digits 18', whereas GNAT on aarch64 only goes up to 64 bits, 'digits
15'.