I have just discovered that convertToWrapped joins the output lines
using LF only, but expects the input to be CRLF.
This seems to be a bug, but means it's not necessary to revert back
after conversion.

It also occurs to me that the correct line-length for wrapping depends
on the output device.

Maybe it would be sensible to set the length to a high number so the
wrapped lines are completely unwrapped?
This would leave the wrapping to the device.

On Wed, 19 Aug 2020 at 12:05, sebb <[email protected]> wrote:
>
> On Wed, 19 Aug 2020 at 09:33, Daniel Gruno <[email protected]> wrote:
> >
> > On 18/08/2020 21.29, sebb wrote:
> > > I wondered about that.
> > >
> > > Now that the reformat is done after MID generation, it would not
> > > affect the generators.
> > > And it would only need to be done for f=f types.
> > >
> > > Also in theory it would only need to be done for mbox imports --
> > > AFAICT live messages *should* have CRLF line terminators.
> > >
> >
> >
> > I've added a LF->CRLF conversion option to the .unflow(), but I have it
> > disabled by default for now, need to figure out the unit tests as they
> > would fail since the resulting body is different from PM <=0.12.
>
> Indeed.
>
> > Perhaps I need to work in some sort of version checks in the unit tests.
>
> I suggest leaving the conversion as an option to make testing easier.
> (It does not have to default to false, as long as it can be overridden)
>
> Most tests would not be affected (f=f is not that common).
> There could be some tests specifically to test the wrapping behaviour;
> these would need to be skipped for earlier versions.
> This could perhaps be done by requiring the dkim generator as an argument?

Reply via email to