[bt-devel] problem with new curl test (Re: Showstopper bugs)

[Eeli Kaikkonen]

On Mon, 9 Nov 2009, Martin Gruner wrote:
> I don't think this is acceptable if it happens regularly.

It's true that letting a segfault just be there forever, every time the
program quits, is not a good option, even if it's harmless. I could of
course report this to Qt and cURL projects and see what they have to
say.

> How much effort would it be to write a custom FTP backend instead?

It's "manageable" but hard to say because it needs some new architecture
and I don't know yet what obstacles I will find.

Eventually we need also new architecture for
threads/installmgrs/progress dialog. The current implemenation creates a
new installmgr for each module and we already have seen it's very
expensive. But it's not only slow but it also eats a huge amount of
memory - hundreds of megs if you try to install a whole repository in
one go. Then the problem of slowness will be seen also when destructing
the objects after all downloads have finished. With my slow laptop it
took much more than 10 minutes IIRC, halting BibleTime meanwhile.

I already have some concrete plans for the transport and architecture,
even code, but it's only a sketch and completely untested.



  Yours,
	Eeli Kaikkonen (Mr.), Oulu, Finland
	e-mail: eekaikko at mailx.studentx.oulux.fix (with no x)