[Edbrowse-dev] curl seg fault

[Edbrowse-dev] curl seg fault

[Karl Dahlke]

If you would pull the latest, and then see if you can reproduce my seg fault,
because it is very very strange.
I think it's another curl thing so hoping Chris can help me here.

First, my latest push fixes a small bug wherein
r filename
would offer to download that file in certain cases.
But if you are commanding to read it into the buffer
then you never want to download it.
That was a one line fix, but did not fix or affect the following bug.

Bring up edbrowse with no args, an empty buffer, and then

r https://login.persona.org/include.js

Type qt to quit, bypassing as much machinery as possible.
seg fault.

This test does not involve javascript at all,
the process is never launched.

The seg fault goes away if you comment out the line
        curl_global_cleanup();
in main.c line 933.

Here's the interesting part; the seg fault also goes away, at least for me,
if you type vs first.
Don't verify certificates.
Read in the file as usual, and quit, and all is well.
What's up with that???
It means if you can't reproduce it it might depend on my local cert file.
It might depend on a lot of things.
Wonder if it is yet another bug in my somewhat older open ssl,
but damnit it's not that much older!

An unrelated question: bgl tracks and displays background downloads,
should it also track and display foreground downloads?
Obviously none of these will ever be in the in progress state.

Karl Dahlke

Re: [Edbrowse-dev] curl seg fault

[Chris Brannon]

Karl Dahlke <eklhad@comcast.net> writes:

> It might depend on a lot of things.
> Wonder if it is yet another bug in my somewhat older open ssl,

I can't reproduce this, go figure.

-- Chris

Re: [Edbrowse-dev] curl seg fault

[Adam Thompson]

[-- Attachment #1: Type: text/plain, Size: 545 bytes --]

On Wed, Jan 14, 2015 at 07:05:46AM -0500, Karl Dahlke wrote:
> Bring up edbrowse with no args, an empty buffer, and then
> 
> r https://login.persona.org/include.js
> 
> Type qt to quit, bypassing as much machinery as possible.
> seg fault.

I can't reproduce this either. I suspect you've just got a broken openssl,
or your version of curl has problems working with your version of openssl.
If I were you I'd seriously consider upgrading.
What distro do you run and do they package an up-to-date openssl and curl?

Cheers,
Adam.

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 473 bytes --]

[Edbrowse-dev] bgl command

[Adam Thompson]

[-- Attachment #1: Type: text/plain, Size: 874 bytes --]

On Wed, Jan 14, 2015 at 07:05:46AM -0500, Karl Dahlke wrote:
> An unrelated question: bgl tracks and displays background downloads,
> should it also track and display foreground downloads?
> Obviously none of these will ever be in the in progress state.

I'd say not. The main reason I asked for this is because when you're browsing
away with something downloading in the background you've got no way of tracking
it and thus don't know when it's ok to quit your browser or what state the
download's in (particularly if you miss the completion or failure message).
If you've downloaded something in the foreground then you'll know when it's
done and probably've done something sensible with it.
I was also wondering if it'd be possible to have a message if you try and quit with a download in progress, like if you quit without writing your changes to a file?

Cheers,
Adam.

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 473 bytes --]