[oi-dev] inkscape calls gcc7 and 10

[Till Wegmüller]

On 07.07.23 07:20, Tim Mooney via oi-dev wrote:

 > The old all-in-one source is no longer a thing, so if we build from
 > released tarballs, we'll potentially need to fetch and extract multiple
 > components.

That has been done before and one component can handle multiple sources. 
Worst case one can use a shellscript to handle the download and placing 
of all sources.

> 1) lib2geom is required, but it could be built and packaged separately.
> Nothing else (other than inkscape) that I'm aware of uses it currently.
> 
> Would it be preferable to have it as a separate package, or just build it
> and package it in the same package as inkscape?
lib2geom would be preferable seperate. Just because it's the first 
consumer it wont mean it's the last.


> 
> 2) some of the extensions require low-use libraries, like the 'libwpd'
> that Udo mentioned.
> 
> Should the inkscape extensions that require these esoteric libraries be
> split into separate packages, so 'image/inkscape' has fewer dependencies,
> but 'image/inkscape/extension/wordperfect' (or whatever) can optionally
> be installed to get support for that legacy format?

Yes spilitting things up is a good idea. A base set can always be 
bundled via requirements via a meta package.

> 
> Next, 1.2.2 actually builds without *any* of the local patches, though I 
> still
> need to look through them to see if there are any OI-customization patches
> that we still need.
> 
> Many of the existing patches are for math library functions like
> log(), sin(), pow(), etc. where integers rather than doubles were being
> passed.  I can update these so they apply to 1.2.2, but they seem to be
> no longer required (probably because of gcc/g++ 10.x).

With int -> double convertions it is simply a build issue.

> 
> Question:
> 
> 3) is it worth updating these patches, if they're not needed to actually
> build any more?  I'm tempted to say no, but I'm willing to update them
> if Andreas or others have strong feelings about keeping as many  > possible.

As Udo mentioned runtime is most problematic.