# Thursday, November 26, 2015¶

## Fixed #624!¶

Yesterday we decided to leave #624 (Ctrl-S in a ComboBox) open because it seems quite hard to solve and because anyway it might vanish after #37, but now I tried a last time to fix it, because it will discourage users from using Ctrl-S. And at least in Chatelet they will have to start using Ctrl-S after the next release if they want to have #557 fixed. In order to solve #557, enter_submits_form must become False. It is possible that in Eupen (not Chatelet) I will have to locally set it to the old value True because in Eupen they don’t have any plain textarea and therefore won’t want to change their habit.

I noticed that it happens only in comboboxes whose store has at least one value to choose from.

It seems that the onLoad was not guilty at all, but after renaming ComboBox.onKeyUp() to deactivate it, the problem disappeared. Of course I cannot simply remove this method because it is important:

onKeyUp : function(e){
var k = e.getKey();
if(this.editable !== false && this.readOnly !== true && (k == e.BACKSPACE || !e.isSpecialKey())){

this.lastKey = k;
}
Ext.form.ComboBox.superclass.onKeyUp.call(this, e);
},


Actually this code is correct, but the problem is that the isSpecialKey() of the Ctrl-S event returned false. So the comboxbox treated Ctrl-S as if it was a normal text entry and accordingly triggered a delayed query of its store.

So I now wrote a hack (see the change in linoweb.js) which makes it return true also for “keyup” events. Hoping that this doesn’t cause any side effect.

Released everything to lino-framework.org so that we can test it on the field.

## AltGr key selects the whole field¶

I finally discovered #631 which had been nagging me already since the last release on lino-framework.org because I cannot enter a @ sign into a Lino application using my Estonian keyboard (I must use S-C-u and type “40” followed by ENTER).

I was fascinated to stumble over the article Why every development team needs continuous delivery. Here is an excerpt:

We already know that continuous integration is the practice of building and running a battery of automated tests after each code change is made or at frequent intervals throughout the day. In other words, continuously (more or less). Continuous delivery adds another step: each time a build passes all your tests, it’s automatically deployed out to a test or staging environment where it can undergo further testing such as load tests, manual exploratory testing, etc. That process may happen just once before changes are released to customers, or you might repeat it for days (or weeks), building up a critical mass of features and fixes, until finally you decide it’s time to ship. (…) Continuous delivery involves human decision-making around what to release to customers, and when. (…) A continuous deployment system automates the entire process from code to customer, allowing you to ship changes several times a day. It’s fabulous in many ways, but it’s just not a good fit for every piece of software. Mobile app users, for example, would go ballistic if they were prompted to download updates every few hours.

Yes, my more than 20 years of experience on the field confirm that frequent deployment is important.

This article was actually an advertisment for JIRA which unfortunately is not free software. But it seems to be good, so Lino Noi might learn from them.

Oh, and how funny: Atlassian are doing the same thing as I have been doing on the Rumma & Ko homepage : they publish their prices for everybody.

And Wikpedia confirms:

Atlassian does not have a traditional sales team. Instead, it lists all prices, information about products, documentation, support requests, and training materials on its website. The company does not offer discounts, with the exception of open source projects, academic and charity organizations. Most of their products are available as hosted or installed versions, starting at \$10 for 10 licenses (pricing does not scale up linearly).

Which actually just confirms that we are on the right way.

## About automated cross-browser layout testing¶

I also read a blog post Is Automated Cross Browser Testing Possible?. It seems that with Selenium it should be possible to write an automated process which takes a list or URLs (I imagine a long list of permalinks to different locations below the online Lino demo applications. The TakesScreenshot class’s (API doc here) docstring has this snippet:

import static openqa.selenium.OutputType.*;

File screenshotFile = ((Screenshot)driver).getScreenshotAs(file);
String screenshotBase64 = ((Screenshot)driver).getScreenshotAs(base64);


Okay this looks promising, but I am not going to use a Java software right now. Actually these things should be done by competent professional partners, IOW services like browserstack.com. So that’s for later.