On 2020/03/08 20:49:46, hahnjo wrote:
> On 2020/03/08 20:45:46, dak wrote:
> > On 2020/03/08 20:25:50, hahnjo wrote:
> > > "sed -i" is not portable and doesn't work on FreeBSD and macOS
IIRC. I
> didn't
> > > check the other commands in here, but I'm sure for that one.
> >
> > Well, maybe, but:
> >
> > git grep 'sed -i'
> > Documentation/GNUmakefile: sed -i -e 's/ISOLANG *= *fr/ISOLANG
=
> > $(ISOLANG)/' $(ISOLANG)/GNUmakefile
> > Documentation/contributor/doc-work.itexi:sed -i -r
> > 's/[0-9a-z]@{40@}/NEW-COMMITTISH/' *.texidoc
> > scripts/auxiliar/update-patch-version.sh:git grep --name-only $1 |
xargs sed
> -i
> > -e s/$1/$2/g
> > stepmake/stepmake/po-targets.make: sed -i '1,2d'
> $(po-outdir)/$(package).po
> > stepmake/stepmake/po-targets.make: sed -i -e 's/^\# This file
is
> > distributed.*/$(sed-header)/' $(po-outdir)/$(package).po
> > stepmake/stepmake/po-targets.make: sed -i -e
's/^\"Content-Type:
> > text\/plain.*/$(sed-content)/' $(po-outdir)/$(package).po
> >
> > At least the stuff in po-targets.make looks like it might get
executed on
> normal
> > builds. Only glanced over the files, though.
>
> Hm okay, never mind then. As a plain build (no docs) is working on
FreeBSD, I
> guess we're not coming across these ;-)
po files are used for LilyPond proper, so if sed -i would be a problem
on FreeBSD, my hunch that those lines would be executed in the context
of a normal build process would be wrong. I am not too thrilled with
the prospect of going through temporary files, and if I change this to
use an ed script, I'd likely get murdered.
On the other hand, the temporary file could just be named $@.tmp without
much of a problem. Right?
https://codereview.appspot.com/567340043/
