[m-dev.] Version of Mercury to Compile Mercury

[Mark Brown]

On Tue, Jun 23, 2015 at 4:50 AM, Zoltan Somogyi
<zoltan.somogyi at runbox.com> wrote:
> On Tue, 23 Jun 2015 01:01:52 +1000, Mark Brown <mark at mercurylang.org> wrote:
>
>> On Tue, Jun 23, 2015 at 12:47 AM, Charles Shuller
>> > I think that isn't recent enough because when I try to configure the code in
>> > git it finds mmc just fine, but still errors out:
>> >
>> > looking for an already installed Mercury compiler to bootstrap with...
>> > checking for mmc... /usr/local/bin/mmc
>> > checking whether the above mmc works and is sufficiently recent... no
>> > configure: error: You need a working Mercury compiler to bootstrap with
>>
>> Make sure this one isn't in the part when you compile the ROTD.
>
> He meant to write "make sure the out-of-date mmc is not in the PATH ...".

Yes. Hopeless spall chucker.

>
> I have been thinking of a way to set things up that would make this simpler.
> We could put a zero-size file into the source distribution (named something
> like ".this_is_a_src_dist" or similar) to act as a flag. The configure script would
> detect this, and ignore any installed mmc, out-of-date or not, to make sure
> that the .c files supplied by the source distribution are not accidentally overwritten.
> The last action of "make install" could be then to delete this file, if it exists.
>
> What do people think?

I agree. Few users would actually benefit from bootstrapping source
distributions from existing installs, and even then they would need to
override that behaviour from time to time.

If you call the file DO_NOT_USE_EXISTING_MMC then it should be easy to
remember which file to remove if the rare case that someone wants to
override the new behaviour.

Cheers,
Mark.