Re: Generate a .pe script for one design size at a time. (issue 93430043)

[hanwenn]

On 2014/05/17 15:13:20, lemzwerg wrote:
> https://codereview.appspot.com/93430043/diff/40001/mf/GNUmakefile
> File mf/GNUmakefile (right):


https://codereview.appspot.com/93430043/diff/40001/mf/GNUmakefile#newcode119
> mf/GNUmakefile:119: true
> Does this really work?  I've never seen a make rule like that, and my
gut
> feeling says that it could be still suffering from race conditions.
Hopefully,
> I'm wrong.

yes, it works.  The previous version still has a problem if the script
creates the SVG before the OTF, so I've added a touch to the dummy
rules, so the timestamps reflect the declared dependencies.

The only situation where this goes wrong is when the .SVG file does not
exist, but the .OTF does, since the recipe doesn't actually produce the
.SVG. Since this is unlikely to occur in practice, I suggest to ignore
this situation.

> The automake info file has a long section w.r.t. commands that produce
multiple
> output files.  Interestingly, your solution with `true' isn't
documented
> there...

this is a variation of the 2nd approach in

http://www.gnu.org/software/automake/manual/html_node/Multiple-Outputs.html

they try to fix it by introducing complex shell which can go wrong in
even more ways.

The really correct is to treat the multiple outputs as one intermediate
file, ie. have the rule creating multiple outputs create a .zip file,
and unpack the final output from the .zip.


https://codereview.appspot.com/93430043/diff/40001/mf/GNUmakefile#newcode136
> mf/GNUmakefile:136: # generating 3 files at the same time
> This comment should be moved up to the first use of `true'.

done.

https://codereview.appspot.com/93430043/