When you specify errorbar linestyles, how do you port them into the legend?

I found a great explanation of how to force a LineStyle change in the errorbars of an errorbar plot, but I've hit a corollary snag in my R2019a environment: when I call legend in my plot using that workaround, the colours and errorbar shapes are ported - but the linestyles I've used the link technique to edit in are not. How can I make the legend reflect my modified errorbars?

h1 = errorbar(1,1,0.3,0.2,'or');
xlim([0 5])
ylim([0 3])
h1.Bar.LineStyle = 'dashdot'
hold on
h2 = errorbar(3,1.4,0.1,0.4,'ok');
h2.Bar.LineStyle = 'dotted'
hold off
hl = legend('VarName1','VarName2');
legend boxoff
title(hl,'LegendTitle')

Note that the figure the code above generates shows different linestyles for the confidence intervals, but that these are not reflected in the legend, which shows solid lines where 'dashdot' or 'dotted' should be used.

I will also note that I can hack in appropriate lines and markers (albeit horiztonal instead of vertical, like the errorbar tokens) by the usual workaround of hold on; plot(desired style but outside desired window) ... ; actual errorbar calls ; hold off. But doing it this way loses the errorbar tokens, which I wanted to keep (assuming the linestyles showed well enough in the vertical space allotted).

[Peter O]

Interesting. Thanks for providing that MWE. I can confirm the issue here, too, on 2019b. I was hoping it might have been a scale thing where the dashes only show up when the legend is bigger so I increased the size of the legend, but the line styles don't change.

Changes to the actual errorbar line's linestyle (e.g. h1.LineStyle=':') fires the event that updates the legend style. I'm wondering if there's a bug buried in the Errorbar class that doesn't register a listener event with the legend for changes to the Bar LineStrip objects. The ErrorBar class seems to live in a protected P-File, which makes it a little tricky to dig around for properties that aren't documented. It might be worth submitting a bug report for your finding!

https://www.mathworks.com/support/bugreports/

[Daniel]

No problem, Peter; MWEs make it so much easier for others to answer! (I also included a sneaky title call in the legend which is technically not MWE, but is something I'm doing in my script. Who knows what subtleties lie under the hood? Hopefully any solution still works with it... and with NumColumns ~= 1, since I'm using that too.) Glad I'm not the only one who can't see how to achieve this.

I'll leave the query open for a little longer before I make a bug report, just in case someone's hit this before but hasn't seen the open query yet.

Thanks for the rapid response :)

[Daniel]

As there have been no further comments/suggestions, I have followed Peter O's advice and reported the bug (#04519315). If I'm sent a workaround, I'll post it here.

[Daniel]

No workaround. Bug noted for future development.

[Adam Danz]

Thanks for the updates, Daniel.

[Daniel]

No problem, Adam. I figure it's far better to be open about the likelihood of a solution here than to leave it otherwise abandoned. That gets my goat when I see it on other questions...

To the community at large: feel free to link to this query if you need an example of an OP proactively responding and keeping the community informed (and incidentally, providing background info + link and a MWE in the post) for someone else.