[m-rev.] for review: fix broken source distribution

[Julien Fischer]

On Sat, 30 Sep 2006, Ian MacLarty wrote:

> Zoltan, I think this should fix the broken source distribution (it's
> still pending bootcheck though).
>
> Another solution to the problem of mtc_union getting compiled with
> debugging might be to use the --force-disable-tracing option in the
> slice and mdbcomp directories.  Do you think that'll work (the library
> will still be compiled with debugging)?
>
> Estimated hours taken: 0.5
> Branches: main
>
> Do not use the installed mmc to compile the slice directory, because
> this means you cannot install the source distribution on a machine that
> doesn't already have Mercury installed.
>
> The reason the slice directory was compiled with the installed mmc was so that
> the mtc_union tool wouldn't be compiled with debugging enabled even if the
> compiler was compiled with debugging.  mtc_union is used to summarise
> trace counts when coverage testing the compiler.  If mtc_union is compiled
> in a debug grade then tail recursion is lost and it runs out of stack space.
>
> If we're going to coverage test the compiler then we should use the
> MERCURY_OPTIONS option that lets us customise the summary command, to choose
> a version of mtc_union not compiled with debugging.
>
> Mmakefile:
> 	Remove PLAIN_SUBDIR_MMAKE and instead use SUBDIR_MMAKE to compile
> 	the slice directory.
>

Another option here would to have configure set the value of
PLAIN_SUBDIR_MMAKE based on whether or not the source distribution is
being compiled.

Julien.
--------------------------------------------------------------------------
mercury-reviews mailing list
Post messages to:       mercury-reviews at csse.unimelb.edu.au
Administrative Queries: owner-mercury-reviews at csse.unimelb.edu.au
Subscriptions:          mercury-reviews-request at csse.unimelb.edu.au
--------------------------------------------------------------------------