The return value of set_url() was not checked, meaning that when it
failed, gmnlm continued anyway, causing an assertion to fail and
subsequentially resulting in a segfault.
---
 src/gmnlm.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/src/gmnlm.c b/src/gmnlm.c
index 41284df..5997edd 100644
--- a/src/gmnlm.c
+++ b/src/gmnlm.c
@@ -780,7 +780,9 @@ main(int argc, char *argv[])
 	}
 
[message trimmed]

Thanks, applied.

Friendly greetings,
Leon Plickat

On Sat Aug 29, 2020 at 8:30 PM CEST, Drew DeVault wrote:
> I would really prefer if you added an automated test for this.

I get it. Would you prefer a test for the next line simply starting
with '\t' or doing a "full" indentation check with parse_indent()?


Friendly greetings,
Leon Plickat

This fixes a bug causing incorrect indentation when the first char in
the next line after an indented line ended with "++" is '\t'.
---
As suggested, I tested this patch with multiple different cases, like
linebreak on indented line and the next line having a (vastly)
different indentation. I also tried generating sway(5) and searching
through that for any unexpected formatting. All cases I tested have
the same result as if generated without this patch, just with correct
indentation after linebreaks. Therefore I still do not think the test
is necessary, so I did not include it in the patch.

 src/main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
[message trimmed]

I just released version 2.0.0 of LavaLauncher.

This is most radical update of LavaLauncher to date. This new version
not only introduces a lot of new features, but also completely
overhauls the configuration as well as basically all internal data
structures and operations. If all patches since the last version are
considered, this is almost a complete rewrite, hence the major version
bump.

CHANGELOG:
- LavaLauncher is now configured via a configuration file instead of
  command flags. This not only provides a considerably more pleasant
  user experience, but also makes more complex configuration options
  possible, allowing LavaLauncher to gain a lot more useful features.

On Wed Aug 26, 2020 at 3:50 PM CEST, Drew DeVault wrote:
> This patch might be a good solution but it would be better with a
> test attached.

I am not sure I understand what test you mean. Test whether the line
is indented before returning and otherwise using the old behaviour?

I may be missing something, but on an unindented line, whether we
directly return after a linebreak or let parse_text() explicitly
handle the newline makes little difference. The resulting output is
the same.

Unrelated, but should the queue be flushed to output before returning?

As the title states, if an indented line is ended with "++" and the
first char on the next line is '\t', indentation in the output is
incorrect.

Input:
```
Line with no indentation.

	Line with indentation++
	and break.
```

Output:
```

Thanks, applied.


Friendly greetings,
Leon Plickat

On Sun Aug 16, 2020 at 5:47 AM CEST, Jan Beich wrote:
> Instead of enumerating each downstream in README consider Repology
> badge. FreeBSD should appear in a few hours, DragonFly (as DPorts)
> after a few weeks.
>
> https://repology.org/project/lavalauncher/badges

Done, thanks your the proposal.


Friendly greetings,
Leon Plickat

On Sun Aug 16, 2020 at 3:32 AM CEST, Jan Beich wrote:
> I've packaged v1.7.1 using -isystem/usr/local/include (USES=localbase)
> as a workaround but master branch required more changes

Is there an official or user curated package I can link in the README?

> May as well upstream, so the rationale isn't lost in time.

Agreed.

I am positively surprised that LavaLauncher works on BSD at all. I can
not promise that I won't accidentally break BSD support in the future,
but I will continue to avoid Linux-isms and GNU-isms (or make them
optional) and will happily merge BSD compatibility patches.