Work on Windows installer

[Richard Kimberly Heck]

On 5/13/20 7:04 AM, Pavel Sanda wrote:
> On Wed, May 13, 2020 at 07:30:51AM +0200, Yu Jin wrote:
>> 1. I could rewrite everything that I think is doable in a much better way,
>> this would mean actually creating a new installer, then checking the old
>> one function by function and then deciding which to rewarite and which to
>> adopt. This would have a potential of new bugs and I would probably share
>> it once it is all finished.
>>
>> 2. Make multiple smaller changes to the current installer always sharing
>> and continuing after they have been accepted. This would take a hell lot of
>> time and require to make sure that the installer still works after each
>> change. The result would probably be not as good as in 1.
>>
>> 3. leaving it as it is and forgetting about it.
>>
>> First option is most appealing to me, well, I probably dont have to tell
>> you why (fun and challenge). If I were to do any work on the installer,
>> then of course I would make sure, that it remains all of it's functionality
>> (e.g. silentinstall) and if I run into problems, I would ask for advice and
>> talk about progress as this would take some time.
>>
>> That said, besides some NSIS experiments I have not done anything on this
>> yet and wanted to talk first to avoid the situation of silently doing
>> something big in my corner without anybody else knowing about it.
>>
>> Please let me know your thoughts.
> Thanks for sharing all these thoughts.

Let me add my thanks, too. This is all old code and a bit of a mess, so
it certainly could use a re-write.


> If you can see better way how to organize the code and improve readability
> that's generally good direction.
>
> If you decide for solution 1:
>
> - the most important point is your willingness to stay around for a while
>   during 2.4.x series and actively fixing installer bugs which inevitably
>   appear after rewrite. This might be sometimes tricky as users are installing
>   lyx on very different setups/windows versions etc.  (i.e. the time for coding
>   per se is only half work you will spend on it at the end...)
>
> - You might want fork the current installer into completely new directory so we
>   have choice to easily switch back if things go wrong.

I think that is an excellent suggestion. Put it in installer2/ or something.


> - Any major rewrite better be finished before we start releasing 2.4 betas so
>   it gets proper testing. JMarc did not share any release plans yet, but my
>   sense is we are talking about couple months(?).

That would be best, if possible.

Riki