Tag Archives: mac

A Windows Developer in Mac Land

This is a guest post from fellow software developer, Simon Kravis.

Few developers would choose their development platform on the merits of their respective Integrated Development Environments (IDEs)  but it happens that applications developed in Windows need to be made available on the Mac platform.

There are many environments offering cross-platform (Mac, Windows and sometimes Android) functionality, but close inspection shows that they all have limitations.  Visual Studio (the native Windows IDE) can produce apps which will run on a Mac using .Net Core – but only if they are command line apps on Windows. Other environments (like Xamarin) do support interfaces, but only involving simple controls like text boxes or drop-downs. There are other cross-platform IDEs (such as Qt)  which offer better graphics support, but they are not cheap and the extent of their support is not evident. If you need functionality such as computer vision, there seems to be no alternative to creating a separate code base for the Mac. Once you start on this path it becomes obvious that Macs handle graphics (and interfaces) very differently from Windows.

Macs have evolved rather more than PCs over the decades: they abandoned their proprietary Mac operating system in favour of UNIX in 1999, adopting the NeXTSTEP platform created by NeXT. Apple originally used PowerPC chips, replacing them with Intel Core processors in 2006, and they are currently transitioning to RISC chips. The Mac NeXTSTEP programming language was Objective C, developed in the 1980s and this is still supported, although the modern Swift language was introduced in 2014, and the Xcode IDE appeared in 2003. Xcode is free, even for teams. It uses the Cocoa API, which is accessible from other environments. The current release (MacOS  13.0) supports both Objective-C and Swift and is also used for developing iPhone and iPad apps. Mac operating systems since Catalina (released in 2019) are 64-bit only.  Xcode can only develop apps for Apple operating systems, notably iOS, which powers the iPhone. Most of the web questions and examples relate to iOS rather than MacOS. MacOS uses different frameworks from iOS, so some functions used in iOS are not available in MacOS, or have different parameters.

The Windows IDE (Visual Studio) dates from 1997, when it bundled together Visual Basic, Visual Fox Pro and Visual Source Safe and Visual C++.  It has an open architecture based on plug-ins and supports 36 different programming languages, but the major ones are C#, VB.Net and C++. Visual Studio can develop apps for any platform via the .NetCore framework, but capability for non-Windows platforms is limited. The Community edition is free, and has almost all the functionality of paid versions.

Both Visual Studio and Xcode are highly complex applications. They both have graphical interface builders where controls are dragged from a library onto a form.  Each application has a vocal supporters and detractors. My experience comes from about 5 years with Visual Studio developing C# applications. Before this I worked with Visual Basic for Applications in Microsoft Access, so I am well-versed in the Microsoft way of doing things.

Like most complex applications, Visual Studio and Xcode each have plenty of bugs, often producing completely unhelpful error messages. Reporting an Xcode bug through standard channels resulted in … nothing. Not even an automated message saying “Thank you for feedback. It will be used to improve future versions”. I haven’t even tried to report a Visual Studio bug, but I suspect that the much larger user base for Visual Studio will mean that workarounds are more readily available, even if the giant ship of Microsoft takes years to respond.

Moving to the Mac and Xcode for development was a shock as I found I didn’t know how to do the most basic things. String manipulation (used in most applications) in Objective C is highly verbose compared to C#. Google was invaluable for finding answers – mostly they were from Stack Overflow, but often from 10 or more years ago, sometimes from Apple Developer Forums. As Xcode has changed considerably since then, answers often had to be adjusted before they could be used.  Another problem is that functionality once provided externally has since been incorporated into Cocoa, so attempts to find a current version of a component (or framework as they called in Cocoa) are often unsuccessful.

MacOS provides more native functionality than Windows. Features such as computer vision and PDF generation are included in MacOS, rather than requiring the use of 3rd party components, which may not as robust as desired, and may require a license for commercial use. However, documentation of MacOS functionality, if present at all, was rarely useful. A few times I asked questions on Stack Overflow which attracted the ire of the Mac gurus for either through having obvious (to them) answers or through not conforming to the forum guidelines (in their opinion). However, the integration of NuGet with Visual Studio provides easy access to the massive number of 3rd party  libraries available for .Net on Windows.

The model-view-controller paradigm used on the Mac took some getting used to, as did the design of the main Xcode screen. Sometimes a useful display would disappear and I had difficulty in finding it how to bring it back. I often had to resort to retrieving earlier versions from the excellent Time Machine backup.  Form design is similar on both platforms – dragging and dropping components from a library. Both Xcode and Visual Studio have bugs, as would be expected for such complex apps. Events from components are generated automatically in Windows, but have to be defined on the Mac (as Actions). References to the component you’ve added also need to be defined on the Mac (as Outlets) and are not a property of the component, whereas on Windows they are.

The Xcode environment provides only basic facilities from scratch: if you need to do something more sophisticated you’ll have to Google around to find out how. Once you know – it’s easy, but the learning curve for Xcode is much higher than for Visual Studio.   

Rather than starting from scratch with the Mac version of my Caption Pro  app, which uses local computer vision functionality to detect multiple photos, changes image dimensions and adds text to images,  I found an existing open-source project on GitHub with similar basic functionality. This dated from 7 years ago and used Objective-C, so that was the language I opted for. An immediate handicap was that many of the answers I found to my questions used Swift in their example code, which is not interconvertible with Objective-C  in the way that C# and VB.Net are. iOS applications for the iPhone (which are most common) use different frameworks from Mac apps, and routines in them sometimes have completely different syntax.

The user interfaces for the Mac and Windows versions look quite different, as shown below. There are some basic differences – menus appear separately to the application window on the Mac and are locked to the top of the screen, whereas Windows menus are part of the application screen. Toolbars offer access to common functionality on the Mac. Differences also arise from the fact the Mac application was adapted from existing code rather than created from scratch.

Figure 1 Windows App main screen
Figure 2 Mac App main screen

Open-source examples (often from GitHub) are useful, but rarely work out-of-the- box. Sometimes the modifications need are minor – like defining the development team-  but sometimes it’s not possible to get them to build in a current version of Xcode.

Debugging on Xcode is frustrating – the call stack frequently contains assembler (which is perhaps why app performance tends to be better on Macs), and the debug variables window does not list all relevant variable values. Variable types may not be correct – Boolean values may appear as dates, and sometimes variables cannot even be evaluated by po (print out) statements. Printing out structure variables may show nothing.  Despite the generally superior performance of Mac apps, building apps in Xcode appears to be much slower than in Visual Studio on similar vintage machines, and after code stops at a breakpoint, it may take a long time before the variables window is filled. Deployment of Mac apps can still be done on an ad-hoc basis, but you have to register as an Apple Developer to avoid blockages in installation arising from being an ‘untrusted source’. Bypassing these blockages is more than a matter of clicking “Install anyway” so it’s hard to avoid forking out US$100 per year for registration. Windows has similar blockages, which can be bypassed with a code-signing certificate. These certificates are available from many vendors, and are slightly cheaper than Apple developer registration, but the process of obtaining one may be very involved.

Ad-hoc deployment is somewhat easier on the Mac than on Windows, but the method of doing it via Archive generation is anything but obvious. Mac applications are actually disk images and applications keep all of relevant files in a folder. This makes uninstallation a matter of dragging the application icon into the recycle bin, a far simpler process than on Windows. dmg files are not recognized by IIS web servers (and may not be by Apache either), so unless the file type is registered, download from a web site will not be possible.

Apple pioneered the App Store for iPhones (it is the only way in which iPhone apps can be installed) and Mac apps can also be put there. Apple takes a commission of 30% (or 15% if you are a small company) and they review all apps before adding them. Passing the review process may be a lngthy process, as not all problems are detected in a review cycle. Fixing these issues and resubmitting may result in further problems coming to light.  The review process may also be somewhat arbitrary. One App Store app presented an interface in German by default. English was available as Preferences option, but only after guessing where the Preferences option was located. App Store apps operate within a sandbox, which places restrictions on filesystem operations. Whether App Store deployment makes economic sense depends on the nature of the app, its market and price structure. Its advantages are that it targets the 16% of desktop users who use Macs, and streamlines installation (and payment, if applicable). The App Store supports ‘freemium’ pricing, where additional features are made available to paying users, but apps with free trial periods are shown as being free but with ‘in-app purchases’, which annoys some users.

Windows deployment can use .msi files, which have been around for decades, but are not easily installed by non-admin users. Self-extracting executables are more tractable, but 3rd party tools have to be used to create them. Windows 10 introduced Universal Windows Programs, which are easier to install and can be placed in the Microsoft Store, which operates in a similar way to the Apple App store, but for Windows desktops and tablets.

A key question which is very difficult to answer is “How long will it take me to convert my Windows app to run on a Mac?” Factors affecting this are app complexity, functionality and programmer skill.  The time between starting work on the Mac app and first deploying it on the company web site was about 3 months, but the amount of time spent on the project each day varied between zero and 3 or 4 hours. If you are a paid resource, then the cost of a cross-platform IDE may be justified, but the requirement for local computer vision functionality added a great deal of complexity to my requirements, which is one reason why I opted for a separate code base. Substantial evaluation would be required before deciding if a cross-platform environment could support any required  functionality.

Simon Kravis runs Aleka Consulting, a small software and consultancy company in Canberra, Australia specializing in information management and offering a number of software products. He has mainly developed scientific and engineering programs, starting in the era of paper tape.

How to add a dark theme to your Qt application

Dark themes are now available for Windows 10 and Mac and it is increasingly expected that desktop applications will offer a dark theme. Previously Qt support for dark themes was patchy. But I am happy to say that it now seems to work fine with Qt 5.12.2, and I have added dark themes to both Windows and Mac versions of my Easy Data Transform and Hyper Plan applications.

Easy Data Transform for Mac with a dark theme:

Easy Data Transform for Windows with a dark theme:

Hyper Plan for Mac with a dark theme:

Hyper Plan for Windows with a dark theme:

I haven’t decided yet whether to add a dark theme to PerfectTablePlan.

Adding dark themes was a fair amount of work. But a lot of that was scouring forums to work out how to integrate with macOS and Windows. Hopefully this article will mean you don’t have to duplicate that work.

Dark themes work a bit differently on Windows and Mac. On Windows changing the UI theme to dark won’t directly affect your Qt application. But you can use an application stylesheet to set the appearance. On Mac changing the UI theme to dark will automatically change your application palette, unless you explicitly block this in your Info.plist file (see below). On both platforms you will need to change any icons you have set to the appropriate light/dark version when the theme changes. Some of this may change in future as dark themes are more closely integrated into Qt on Windows and Mac.

macOS

You can add the following helper functions to a .mm (Objective-C) file:

#include "Mac.h"
#import <Cocoa/Cocoa.h>

bool macDarkThemeAvailable()
{
    if (__builtin_available(macOS 10.14, *))
    {
        return true;
    }
    else
    {
        return false;
    }
}

bool macIsInDarkTheme()
{
    if (__builtin_available(macOS 10.14, *))
    {
        auto appearance = [NSApp.effectiveAppearance bestMatchFromAppearancesWithNames:
                @[ NSAppearanceNameAqua, NSAppearanceNameDarkAqua ]];
        return [appearance isEqualToString:NSAppearanceNameDarkAqua];
    }
    return false;
}

void macSetToDarkTheme()
{
   // https://stackoverflow.com/questions/55925862/how-can-i-set-my-os-x-application-theme-in-code
   if (__builtin_available(macOS 10.14, *))
   {
        [NSApp setAppearance:[NSAppearance appearanceNamed:NSAppearanceNameDarkAqua]];
   }
}

void macSetToLightTheme()
{
    // https://stackoverflow.com/questions/55925862/how-can-i-set-my-os-x-application-theme-in-code
    if (__builtin_available(macOS 10.14, *))
    {
        [NSApp setAppearance:[NSAppearance appearanceNamed:NSAppearanceNameAqua]];
    }
}

void macSetToAutoTheme()
{
    if (__builtin_available(macOS 10.14, *))
    {
        [NSApp setAppearance:nil];
    }
}

The macSetToLightTheme() and macSetToDarkTheme() are useful if you want to give the user the option to ignore the OS theme. Call macSetToAutoTheme() to set it back to the default.

Corresponding header file:

#ifndef MAC_H
#define MAC_H

bool macDarkThemeAvailable();
bool macIsInDarkTheme();
void macSetToDarkTheme();
void macSetToLightTheme();
void macSetToAutoTheme();

#endif // MAC_H

You then need to add these files into your .pro file:

macx {
   ...
   HEADERS += Mac.h
   OBJECTIVE_SOURCES += Mac.mm
}

You can detect a change of theme by overriding changeEvent():

void MainWindow::changeEvent( QEvent* e )
{
#ifdef Q_OS_MACX
    if ( e->type() == QEvent::PaletteChange )
    {
        // update icons to appropriate theme
        ...
    }
#endif
    QMainWindow::changeEvent( e );
}

If you decide you *don’t* want to add a dark theme to your Mac app, the you should add the bold entry below to your Info.plist file:

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
    ...
    <key>NSRequiresAquaSystemAppearance</key>
    <true/>
</dict>
</plist>

This will then force it to be shown in a light theme, regardless of the theme the operating system is in.

Windows

To set a dark theme palette you can use a stylesheet:

QFile f( ":qdarkstyle/style.qss" );
if ( !f.exists() )
{
   qWarning() << "Unable to set dark stylesheet, file not found";
}
else
{
   f.open( QFile::ReadOnly | QFile::Text );
   QTextStream ts( &f );
   getApp()->setStyleSheet( ts.readAll() );
}

The stylesheet I used was a modified version of qdarkstyle from a few years ago.

To unset the stylesheet and return to a light theme just call:

getApp()->setStyleSheet( "" );

Alternatively you can do it by changing the application palette.

Windows helper functions:

bool windowsDarkThemeAvailable()
{
    // dark mode supported Windows 10 1809 10.0.17763 onward
    // https://stackoverflow.com/questions/53501268/win10-dark-theme-how-to-use-in-winapi
    if ( QOperatingSystemVersion::current().majorVersion() == 10 )
    {
        return QOperatingSystemVersion::current().microVersion() >= 17763;
    }
    else if ( QOperatingSystemVersion::current().majorVersion() > 10 )
    {
        return true;
    }
    else
    {
        return false;
    }
}

bool windowsIsInDarkTheme()
{
    QSettings settings( "HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Themes\\Personalize", QSettings::NativeFormat );
    return settings.value( "AppsUseLightTheme", 1 ).toInt() == 0;
}

Currently there seems to be no way to conect to a signal or event that shows the theme has changed in Windows. So I connected to a signal from a QTimer that fires every 5 seconds to check windowsIsInDarkTheme().

Icons

When the theme changes you potentially need to update any icons you have set, e.g. for the toolbar.

In a light theme you can usually set the active icons and let Qt calculate the corresponding disabled icons. This doesn’t work for a dark theme as you want the disabled icons to be darker than the enabled icons, rather than lighter. So you can either calculate the disabled icons programmatically or you can provide a set of disabled icons as well. I opted for the former.

Assuming your icons are set up as resources under :/icons/dark and :/icons/light you can do something like this:

QString getResourceName( const QString& iconName, bool dark )
{
    return QString( ":/icons/%1/%2" ).arg( dark ? "dark" : "light" ).arg( iconName );
}

QPixmap getPixmapResource( const QString& iconName, bool dark )
{
    QString resourceName = getResourceName( iconName, dark );
    QPixmap pixmap = QPixmap( resourceName );
    Q_ASSERT( !pixmap.isNull() );
    return pixmap;
}

QIcon getIconResource( const QString& iconName, bool dark )
{
    QIcon icon;
    QPixmap pixmap = getPixmapResource( iconName, dark );
    icon.addPixmap( pixmap );
    if ( dark )
    {
        // automatic disabled icon is no good for dark
        // paint transparent black to get disabled look
        QPainter p( &pixmap );
        p.fillRect( pixmap.rect(), QColor( 48, 47, 47, 128 ) );
        icon.addPixmap( pixmap, QIcon::Disabled );
    }
    return icon;
}

Then you can reset the icon for the appropriate theme with:

bool isDark()
{
#ifdef Q_OS_MACX
   return macIsInDarkTheme();
#else
   return windowsIsInDarkTheme();
#endif
}
...
myButton->setIcon( getIconResource( "my_icon.png", isDark() ) );
...

You may also be able to update icons through QIcon::setThemeName(). But I didn’t explore this in any detail.

Note that you probably don’t want the enabled icons to be pure white, as it is a bit too visually jarring against a dark theme.

Running Qt apps on M1 ARM Macs

Apple is switching the processor architecture of it’s Macs. Again (I transitioned PerfectTablePlan from PowerPC to Intel some hears ago). This time to their own M1 ARM chips. Reports so far have been very positive about speed and battery life of the new processors. Obviously most current Mac software has been written for Intel Macs, so they are using the Rosetta2 emulation layer to run apps compiled for Intel Macs on the ARM chips. I’m not sure how much of a performance hit this causes, but clearly it would be better to run native ARM binaries on an ARM machine. Also Apple, being Apple, want to move everyone to ARM as quickly as possible. Tough luck if you just spent big bucks on a shiny new Intel Mac.

One of my customers emailed me that the latest version of my Hyper Plan visual planner, built with Qt 5.13.1, didn’t run on an new M1 Mac. I don’t currently have an M1 Mac to test it on. But my Easy Data Transform software , built with Qt 5.15.2, apparently works fine on an M1 Mac. So I recompiled Hyper Plan using Qt 5.15.2, and was told it now works. I have found a couple of minor differences in behaviour between Qt 5.13.1 and 5.15.2, but they are too obscure to go into here. Some Qt apps may still have issues on ARM.

Currently Qt is only available as Intel binaries. Efforts are in progress to be able to build Qt as M1 (ARM) binaries. When that is complete it should be possible to ship Qt applications as a ‘fat binary’ with both Intel and ARM executables, as I did with the PowerPC to Intel transition. I’m not sure if this is going to be supported on Qt 5 and 6 or just Qt 6.

** Update Dec-2021 **

Qt 6.2 supports building M1 ARM and Intel binaries. There is no official support for M1 Arm binaries for Qt 5.

** Update Apr-2022 **

Evan of ModernCSV alterted me to an article on deploying a ‘fat binary’ from Qt 5.

Bloviate

I wondered what it would look like if you took a body of text and then used it to generate new text, using Markov chains of different lengths. So I knocked up  quick program to try it.  ‘Bloviate’.

bloviate

Bloviate analyses your source text to find every sequence of N characters and then works out the frequency of characters that come next.

For example, if you set N=3 and your source text contains the following character sequences staring with ‘the’:

‘the ‘, ‘then’, ‘they’, ‘the ‘

Then ‘the’ should be followed 50% of the time by a space, 25% of the time by an ‘n’ and 25% of the time by a ‘y’.

Bloviate then creates output text, starting with the first N characters of the source text and filling in the rest randomly using the same sequence frequencies as the source text.

Note that a character is a character to Bloviate. It treats upper and lower case as different characters, makes no attempt to differentiate between letters, punctuation and white space and does not attempt to clean up the source text. Which also means it works on any language.

Bloviate also tells you the average number of different characters following each unique sequence of N, which I will call F here. As F approaches 1.0 the output text becomes closer and closer to the input text.

Using ‘Goldilocks and the 3 bears’ as input:

If N=1 (F=7.05) the output is garbage. Albeit garbage with the same character pair frequency as the original.

On cre She sl s ramy raked cheais Bus ore than s sherd up m. ged. bend staireomest p!”Sof ckstirigrorr a ry ps.

” f waine tind s aso Sowa t antthee aime bupis stht stooomed pie k is beche p!

At N=3 (F=1.44) it looks close to English, but jibberish:

Once up and been sight,” she this timed. Pretty so soon, she second soft. She screame up and she screame hot!” cried the Mama bed the Papa been sleeping in the Papa bear

“Someone’s bear growl.

At N=5 (F=1.14) it starts to look like proper English, but semantically weird:

Once upon a time, so she went for a walked right,” she lay down into the kitchen, Goldilocks sat in the porridge from the three chair,” growled, “Someone’s been sitting my porridge and she tasted the door, and ran down the bedroom. Goldilocks woke up and she second bowl.

And it comes out with occasional gems such as:

“Someone’s been sitting my porridge,” said the bedroom.

At N=10 (F=1.03) it starts to become reasonably coherent:

Once upon a time, there was a little tired. So, she walked into the forest. Pretty soon, she came upon a house. She knocked and, when no one answered, she walked right in.

At the table in the kitchen, there were three bowls of porridge.

At N=15 (F=1.01) it starts to get pretty close to the original text, but doesn’t follow quite the same order:

Once upon a time, there was a little girl named Goldilocks. She went for a walk in the forest. Pretty soon, she came upon a house. She knocked and, when no one answered, she walked right in.

At the table in the kitchen, there were three bowls of porridge. Goldilocks was very tired by this time, so she went upstairs to the bedroom. She lay down in the first bed, but it was too hard. Then she lay down in the third bed and it was just right. Goldilocks fell asleep.

At N=12 (F=1.07) the whole 680k characters of ‘Pride and prejudice’ produces:

It is a truth universally contradict it. Besides, there was a motive within her of goodwill which could not help saying:

“Oh, that my dear mother had more command over herself! She can have her own way.”

As she spoke she observed him looking at her earnest desire for their folly or their vice. He was fond of them.”

Obviously the source text is important. The Bohemian Rhapsody lyrics make nearly as much (or as little sense) at N=5 (F=1.08) as the original:

Is this to me, for me, to me

Mama, just a poor boy from this to me

Any way the truth

Mama, life? Is this time tomorrow

Carry on as if nothing all behind and face the truth

Mama, ooh, didn’t mean to me, baby!

Just gotta leave me and lightning, very fright out, just killed a man

Put a gun against his head

Pulled my time to die?

At N=12 (F=1.05) 160k characters of Trump election speeches produces:

Hillary brought death and disaster to Iraq, Syria and Libya, she empowered Iran, and she unleashed ISIS. Now she wants to raise your taxes very substantially. Highest taxed nation in the world is a tenant of mine in Manhattan, so many great people. These are people that have been stolen, stolen by either very stupid politicians ask me the question, how are you going to get rid of all the emails?” “Yes, ma’am, they’re gonna stay in this country blind. My contract with the American voter begins with a plan to end government that will not protect its people is a government corruption at the State Department of Justice is trying as hard as they can to protect religious liberty;

Supply your own joke.

I knocked together Bloviate in C++/Qt in a couple of hours, so it is far from commercial quality. But it is fairly robust, runs on Windows and Mac and can rewrite the whole of ‘Pride and prejudice’ in a few seconds. The core of Bloviate is just a map of the frequency of characters mapped to the character sequence they follow:

QMap< QString, QMap< QChar, int > >

You can get the Windows binaries here (~8MB, should work from Windows 7 onwards).

You can get the Mac binaries here (~11MB, should work from macOS 10.12 onwards).

Note that the Bloviate executable is tiny compared to the Qt library files. I could have tried to reduce the size of the downloads, but I didn’t.

To use Bloviate just:

  1. paste your source text in the left pane
  2. set the sequence length
  3. press the ‘Go >’ button

I included some source text files in the downloads.

You can get the source for Bloviate here (~1MB).

It should build on Qt 4 or 5 and is licensed as creative commons. If you modify it, just give me an attribution and send me a link to anything interesting you come up with.

How to notarize your software on macOS

** Please note: WordPress keeps mangling my code examples by changing double dash to single dash. I tried to fix it. But it changed them all back again! If anyone knows how to get around this, please put something in the comments. **

Apple now wants you to ‘notarize’ your software. This is a process where you upload your software to Apple’s server so it can be scanned and certified malware free. This will probably become compulsory at some point, even (especially?) if your software isn’t in the Apple app store. Apple says:

Give users even more confidence in your software by submitting it to Apple to be notarized. The service automatically scans your Developer ID-signed software and performs security checks. When it’s ready to export for distribution, a ticket is attached to your software to let Gatekeeper know it’s been notarized.

When users on macOS Mojave first open a notarized app, installer package, or disk image, they’ll see a more streamlined Gatekeeper dialog and have confidence that it is not known malware.

Note that in an upcoming release of macOS, Gatekeeper will require Developer ID signed software to be notarized by Apple.

Documentation on notarization is a bit thin on the ground, especially if you want to notarize software that wasn’t built using XCode (I build my software using QtCreator). So I am writing up my experiences here.

First you need to ensure you have macOS 10.14 and XCode 10 installed (with command line tools) and you need a current Apple developer account.

Codesign your app with ‘hardened runtime’ using –options runtime :

codesign –deep –force –verify –verbose –sign “Developer ID Application:<developer id>” –options runtime <app file>

E.g.:

codesign –deep –force –verify –verbose –sign “Developer ID Application: Acme Ltd” –options runtime myApp.app

A ‘hardened runtime’ limits the data and resourced an application can access. I’m not sure what the exact ramification of this are. But it doesn’t seem to have restrict my software from doing anything it could do previously.

You can check the signing with:

codesign –verify –verbose=4 <app file>

E.g.:

codesign –verify –verbose=4 myApp.app

Now package your app into a .dmg (e.g. using DropDMG). Then upload the .dmg to Apple’s servers:

xcrun altool -t osx -f <dmg file> –primary-bundle-id <bundle id> –notarize-app –username <username>

E.g.:

xcrun altool -t osx -f myApp.dmg –primary-bundle-id com.acme.myapp –notarize-app –username me@acme.com

You will be prompted for your Apple developer password (or you can include it on the command line).

You now have to wait a few minutes. If the upload is successful “No errors uploading ” will be shown and a unique ID will be returned. You then have to use this to request your upload be scanned:

xcrun altool –notarization-info <notarize ID> -u <username>

E.g.:

xcrun altool –notarization-info xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxx -u me@acme.com

You will be prompted for your Apple developer password (or you can include it on the command line).

Hopefully you will see “Status Message: Package Approved”. If the notarization fails, you should be sent a link to an online log file describing the issue. If the notarization completes successfully you need to ‘staple’ the results to your .dmg:

xcrun stapler staple -v <dmg file>

E.g.:

xcrun stapler staple -v myApp.dmg

The stapler outputs a log including some odd phrases. Mine included: “Humanity must endure”, “Let’s see how that works out. “, “Adding 1 blobs to superblob. What about Blob?” and “Enjoy”. Weird. Hopefully it will end with “The staple and validate action worked!”.

Finally you can unpack your .dmg into a .app and verify it with:

spctl -a -v <app file>

E.g.

spctl -a -v /Applications/myApp.app

On macOS 10.14 (but not earlier OSs) it should say “source=Notarized Developer ID”. Your software should now run on 10.14 without a warning dialog. Congratulations!

It all seems rather clumsy. As you have to wait asynchronously for the unique ID to be returned from step 1 before you can complete step 2, it is not easy to fully automate in a script. This is a major pain the arse. If anyone works out a way to automate it the whole process, please let me know.

Here are some links to the various posts that I gleaned this information from:

https://cycling74.com/forums/apple-notarizing-for-mojave-10-14-and-beyond
https://www.mbsplugins.de/archive/2018-11-02/Notarize_apps_for_MacOS
https://forum.xojo.com/50655-how-to-codesign-and-notarise-your-app-for-macos-10-14-and-highe
https://forum.xojo.com/49408-10-14-hardened-runtime-and-app-notarization/11
https://stackoverflow.com/questions/53112078/how-to-upload-dmg-file-for-notarization-in-xcode
https://lapcatsoftware.com/articles/debugging-mojave.html

Getting Qt 5.9 working on Windows (eventually)

I have had Qt 5.5 and 5.6 installed on my development machines for some time. Now that I have purchased a new Mac development box (an iMac with a lickably beautiful 27″ screen) I thought it was a good time to update to a more recent version of Qt. I went for Qt 5.9, rather than Qt 5.10, as 5.9 has been designated as an LTS (long term support) release. Upgrading turned into a real chore. I am quickly writing it up here in the hope that it helps someone else, and as a reminder to myself a few years down the line.

I like to build Qt from source. Because then I know it was built using the same compiler, headers, SDK etc as I am using to build my product. And I have more control over how Qt is configured. Also I can patch the source and rebuild it, if I need to. But I have had problems building Qt on Mac before. So I decided to install the pre-built binaries on my new Mac. I installed the latest version of XCode and then the Q5.9.4 binaries. This was a couple of big downloads, but it all went pretty smoothly.

I successfully built Qt 5.5 from source on my Windows machine previously, so I decided to try that for Qt 5.9. I have Visual Studio 2010 installed. This isn’t supported for Qt 5.9.4, so I downloaded Visual Studio 2017. I unzipped the Qt source into C:\Qt\5.9.4, ran ‘x86 native tools command prompt for VS 2017’, made sure Python and Perl were in the path and then:

cd C:\Qt\5.9.4

set QTDIR=C:\Qt\5.9.4\qtbase

set PATH=%QTDIR%\bin;%PATH%

configure -opensource -confirm-license -opengl desktop -nomake tests -nomake examples -no-plugin-manifests -debug-and-release -platform win32-msvc -verbose

nmake

Note that you are told by the nmake script to do nmake install at the end of this. But it tells you somewhere in the Qt Windows documentation not to do this, unless you have set the prefix argument (confusing, I know)

The build failed part way through making qtwebengine. Something to do with a path being too long for Perl or Python (I forget). It seems to be a known problem. Odd as the root path was just C:\Qt\5.9.4. I don’t need qtwebengine at present, so I deleted everything and tried again with -skip qtwebengine:

configure -opensource -confirm-license -opengl desktop -skip qtwebengine -nomake tests -nomake examples -no-plugin-manifests -debug-and-release -platform win32-msvc -verbose

nmake

It seemed to complete ok this time. But using this version of Qt to build Hyper Plan I got an error:

Unknown module(s) in QT:svg

On further examination the SVG DLL  had been built, but hadn’t been copied to the C:\Qt\5.9.4\qtbase\bin folder. Similarly for a lot of the other Qt DLLs. I couldn’t find any obvious reason for this looking through logs, Stackoverflow and Googling. I could possibly do without the SVG functionality, but I wasn’t sure what else was broken. So I decided to give up on bulding from source on Windows as well.

I download the Qt 5.9.4 binaries for Visual Studio 2017. This seemed to go ok, but then I discovered that I could only build a 64-bit application from these. No 32-bit version was available for Visual Studio 2017. Many of my customers are still on 32 bit versions of Windows. So I need to be able to ship my product as a 32 bit executable + DLLs[1].

So I uninstalled Visual Studio 2017 and installed Visual Studio 2015. I then got an error message about Visual Studio 2017 redistributables that I hadn’t uninstalled. So I had to uninstall those and run a repair install on Visual Studio 2015. That seemed to work ok. So then I download the 32-bit Qt 5.9.4 binaries for Visual Studio 2015. I had to download these into a different top level folder (C:\Qtb), so as not to risk wiping existing Qt installs that I had previously managed to build from source.

Eventually I managed to build Hyper Plan and PerfectTablePlan on Mac and Windows. What a palaver though! Qt is an amazing framework and I am very grateful for everyone who works on it. But I wish they would make it a bit easier to install and upgrade! Has anyone actually managed to get Qt 5.9 built from source on Windows?

[1] I don’t bother shipping a 64-bit executable on Windows as the 32-bit executable works fine on 64-bit versions of Windows (my software doesn’t require excessive amounts of memory). I only ship a 64-bit executable on macOS as almost no-one uses 32-bit versions of macOS now.

Bundlefox review

I have been using bundles and 1-day sales as a useful way to increase the exposure for my visual planning software. I have had positive experiences with BitsDuJour, Macupdate and BundleHunt. Once you put your software in one bundle you inevitably get approached by people who run other bundle promotions. I was approached by Bundlefox and agreed to put Hyper Plan in their Mac software bundle. I wish I hadn’t. It has been a pretty miserable experience from start to finish. In brief:

  • I never knew when the promotion was going to start or end. I was told it was going to start on 27th February, but it eventually started on 20th April. It was supposed to run for 3 weeks, but actually ran for 6 weeks. This is a problem, because it means you can’t put your software in other sale or bundle that require an exclusive discount.
  • Communication was poor. They generally took several days to reply to emails.
  • The number of licenses sold was very low, especially compared with sales of Hyper Plan on BundleHunt.
  • Worst of all, they only paid me 60% of what I was expecting per license. When I queried this they emailed me back “It’s **% revenue share after fees, most of the sales came in through affiliates and we had to pay them off before sharing the revenue”. I went back through their emails and their ‘Vendor Manual’ and there is no mention of affiliate fees being subtracted. It just says “You would receive a percentage of the total payments received for the bundle minus PayPal fees”. In fact I had emailed them “So if you sell 2000 bundles for $12 of which 500 choose Hyper Plan, I get **% of $12×500 = $***?” and they replied “Your calculation is correct”. I feel deceived.
  • The low number of licenses sold and the low payout per license means that it wasn’t worth the effort to setup.

I don’t know what Bundlefox are like to deal with as buyer, but I recommend vendors give them a wide berth.

Promoting your software through 1-day sales and bundles

Hyper Plan, my visual planning software for Windows and Mac, has now been for sale for a bit less than 2 years. Given that I am (by choice) doing all the development, marketing and support for both Hyper Plan and my other product, PerfectTablePlan, I have had a limited amount of time to promote Hyper Plan. But Hyper Plan is in a  competitive market, where it is hard to get noticed using traditional promotional techniques such as SEO and PPC. So I have been experimenting with promotion via 1-day sales sites and bundles.

I did several promotions through both bitsdujour.com and macupdate.com promo. These were 50%-off sales for 1 day (sometimes extended for another day). The site takes 50% commission on the sale, so I only got $10 of my normal $40 ticket price. But I also got exposure to a whole new audience I wouldn’t normally reach.

I also included Hyper Plan in bundlehunt.com and macupdate.com software bundles. In these bundles customers purchased some 10 items of software at a big discount. The promotions lasted for a few weeks each. I am not at liberty to divulge how much I got for each licence, but a quick calculation based on the price of the bundles and the number of items in the bundle tells you that it was a lot less than $10!

My hopes related to sales sites and bundles were:

  1. A worthwhile amount additional sales revenue.
  2. Increased feedback, giving me more insight for improving the product.
  3. Making money further down the line from major upgrades (e.g. v1 to v2).
  4. That I wouldn’t be swamped in support emails from people who were paying me a lot less than the standard price.
  5. More word-of-mouth sales after the discount has finished.

On analysing the results, the first 4 turned out to be true.

I had previously tried promoting my PerfectTablePlan table planning software on bitsdujour.com, but the results were disappointing. It just wasn’t a good match for their audience. However Hyper Plan is a more general tool and it did a lot better. The bundles also sold in impressive volumes. The source of Hyper Plan sales revenues to date after commission (but not including upgrades) is show below.

sales-revenue-source

So the extra sales were certainly significant from a revenue point of view, bearing in mind that Hyper Plan is a relatively young and unknown product.

I also got some very useful feedback from the bitsdujour comments section.

I released v2 of Hyper Plan in March 2016. I have crunched the numbers to see how many v1 customers to date have paid for upgrades to v2.

percentage upgrades

I expected that the 1-day sale customers who had paid $20 for the initial licence would be less likely to pay $16 to upgrade to v2 than those who had hadn’t purchased at a heavy discount. I was surprised that the opposite turned out to be true. I don’t have a good theory why.

I don’t have any figures for bundle customer upgrades, as the bundles happened after v2 was released. Given that bundle purchasers probably only wanted a subset of the software in the bundle, I expect the upgrade percentages to be a lot lower than above.

I wasn’t swamped in support emails. In fact things were surprisingly quiet during the bundles, which makes me wonder how many people who purchased the bundle were interested in Hyper Plan.

There were no sustained jumps in traffic or sales after the 1-day sales or bundles ended.

Best of all, the 1-day sales and bundles don’t cost anything, apart from a modest amount of time to set-up.

I know some vendors promote these 1-day sales and bundles to existing customers. But I don’t understand why you would do that. The whole point of these channels is to reach new audiences. Also you risk annoying customers who have paid list price. If you already have an audience you can promote a sale to, then you don’t need 1-day sales sites or bundles. Just email them a discount voucher.

I had one complaint from an existing customer on a forum who had paid full price and then saw Hyper Plan in a 1-day sale. I offered to refund the difference back to them, but they didn’t take me up on it.

In conclusion, the sales and bundle sites brought in useful spikes of additional sales (especially when you include upgrades later on) and feedback, without a big jump in support burden. But they didn’t lead to a noticeable long-term increase in traffic or sales. Obviously every product is different. But if you have a product that needs exposure, isn’t too niche and doesn’t require a lot of support, it may be worth giving 1-day sales and bundles a try.

Hammer For Mac static website generator

I prefer static websites to a CMS for simple product websites because:

  • Static websites are fast.
  • I have more low-level control over the HTML/CSS.
  • I don’t have to worry about the very-real threat of a CMS being hacked.

Obviously writing every page separately in raw HTML/CSS would go against one of the cardinal rules of development, Don’t Repeat Yourself. But you can avoid this using a static website generator such as Hammer for Mac.

hammer

Hammer uses a simple syntax embedded in HTML comments to ‘compile’ a website from source files. I have now used Hammer to create several static HTML/CSS websites, including my perfecttableplan.com and hyperplan.com websites.

I like the simple syntax of Hammer. For example:

I can put the HTML for a page header in an _header.html file and then each page just needs to start with:

<!-- @include _header.html -->

I can define and use variables:

<!-- $current_year 2016 -->
..
<p>Copyright <!-- $current_year -->.</p>

And I can let Hammer work out relative paths:

<img src="@path image.png" />

If Hammer can’t make sense of a source file (e.g. it can’t find the image file), it generates a compilation error.

Because everything is text based I can easily manage all the source in a version control system. Also, if I have to move away from Hammer, it should be relatively straightforward to change the syntax to another static generator (or even write a replacement for Hammer!).

Overall I like Hammer. But it does have a number of shortcomings:

1. The user interface is very limited. Hammer shows you a list of source files and you can click on a source file to see the compiled version or edit the source. But the source files are listed in the order they were edited and you can’t filter or sort the list. This seems such a simple and basic feature, that I can’t understand why the developers have omitted it.

2. Hammer takes a dumb, brute force approach to compilation. If you change any file in a source folder, it recompiles *everything*, without checking if other source files include that file. This is a pain if you have 100+ source files. Surely it wouldn’t be that hard to work out which files depend on which and only recompile the files that need recompiling?

3. You can’t nest variables. For example you can’t do this:

<!-- $current_year 2016 -->
<!-- $copyright_message Copyright <!-- $current_year --> -->

This might sound minor. But it limits the expressiveness of variables significantly.

4. The vendor doesn’t do email support. If you want to communicate with them you have to use Slack or Twitter. I am old fashioned, I like email.

5. It only runs on Mac OS X (the clue is in the name).

At one point Hammer looked like abandonware, but owner riothq.com sold it to beach.io and active development has resumed.

Currently Hammer is priced at £15.39 (and presumably some round number of US dollars). That seems way too cheap. I wish they would price it a bit higher and fix some of the issues above.

** Update Jan-2022 **

Hammer4mac doesn’t run on recent versions of macOS and seems to have been quietly abandoned by the developer for some time.

South West Bootstrappers meetup

I am organizing a regular meetup in Swindon (UK) for people who are running (or are interested in running) their own bootstrapped (i.e. not VC funded) software product business. Come along and talk shop with other aspiring and experienced bootstrappers. It doesn’t matter if you are developing for web, Windows, Mac or mobile.

The first meetup is on the evening of Tuesday 16th June 2015. You can find out more and RSVP at meetup.com/South-West-Bootstrappers/.

swindon meetup