Re: Building LO fails with autogen

It's the exact same document in both cases, and the conversion is
auto-piloted using Python and UNO.  In fact, it's the exact same
scenario on both hosts:

  - run LO headless,
  - open Python interpreter and use UNO to connect to LO,
  - iterate over all paragraphs.

Everything (LO, Python, ...) runs local on the host, yet there are these
huge performance differences: 3 sec on Mac, ~3 min on Linux.

I have posted a question but there's no response:

  http://en.libreofficeforum.org/node/9850

Cheers,
Jens


On Sat, Dec 27, 2014 at 07:17:48PM +0100, Miklos Vajna wrote:
> The conversion time has a constant part (LO starting up) and a variable
> part that depends on the import/export filters. For simple input, a
> conversion takes about 2 seconds for me, so 3 minutes definitely sounds
> like a bug, unless your input is some benchmark document. ;-)
>
> If you do some mass-conversion and the 2 seconds constant overhead isn't
> acceptable, then have a look at using LO via LOK or UNO that allows
> multiple conversions with a single startup.
>
> All in all, --without-x (formerly known as --enable-headless) probably
> will just bring you more troubles, as almost nobody uses that option, so
> you will start using some code path that gets very little testing.

-- 
Jens Tröger
http://savage.light-speed.de/