Re: cygwin64 - building 2.18.2 doc fails

[David Kastrup]

Masamichi HOSODA <address@hidden> writes:

> I've attach these patches to this mail.
>
>   cygwin-python.patch:
>     Set LDFLAGS to build python module
>
>   cygwin-remove-pathconv.patch:
>     Remove cygwin_conv_to_posix_path
>     This patch is similar to yours.

Don't know enough about Cygwin to know whether those two are the right
thing.  I'd recommend passing them through review separately.

>   cygwin-env-TMP.patch:
>     Don't use environment variable TMP when "make doc"
>     If I understand correctly,
>     cygwin system use environment variable TMP for a special purpose.
>     In other words, on cygwin,
>     it can not be used for other purpose like Makefile variable.

That one is clearly a good idea.  I'd not mind it being pushed directly
to staging.

However, doing "git grep TMP" I also find
scripts/lilypond-invoke-editor.scm:                                (or (getenv 
"TMP")

which is from

                                   (or (getenv "TMP")
                                       (getenv "TEMP")
                                       "/tmp")

and

stepmake/stepmake/metafont-rules.make:  $(DO_MF_DEP) TMP=`mktemp -d 
$(outdir)/pfbtemp.$*.XXXXXXXXX` \
stepmake/stepmake/metafont-rules.make:  && ( cd $$TMP \
stepmake/stepmake/metafont-rules.make:  && mv $$TMP/*pfb $(outdir); \
stepmake/stepmake/metafont-rules.make:  rm -rf $$TMP

The latter would need another similar fix.  And the former uses
Windows-only conventions: the POSIX convention is to look in the TMPDIR
variable.  So it might warrant putting a TMPDIR check in front.

The TMPDIR patch should likely be separate (as it is a different issue
and just discovered in parallel).  But the metafont-rules.make fix
belongs with the other ones I think.

-- 
David Kastrup