Skip to content

Latest commit

 

History

History
217 lines (136 loc) · 15.2 KB

README.md

File metadata and controls

217 lines (136 loc) · 15.2 KB

iOS Crash Dump Analysis Book

The HTML version of the book, "iOS Crash Dump Analysis" is available as follows:

Language Link
English https://faisalmemon.github.io/ios-crash-dump-analysis-book/en
Chinese https://faisalmemon.github.io/ios-crash-dump-analysis-book/zh

Accompanying the book is an Analytic Troubleshooting worksheet. Download the Analytic Troubleshooting worksheet pdf

The Kindle version is available for purchase at: Amazon.com, Amazon.co.uk, and other geographic regions of Amazon.

The paperback print version is available for purchase at: Amazon.com, Amazon.co.uk, and other geographic regions of Amazon.

Note, as an Amazon Associate, I earn from qualifying purchases made via the above links.

The print edition is also from traditional book sellers, libraries, and other institutions.

The purpose of this repository is to provide the source code, and other resources, supporting the iOS crash dump analysis book.

The rationale for providing the source freely, in addition to the commercial offering, is to make it easy to collaborate, share, and build upon the ideas of the book. Buying a print, or electronic, version of the book helps support the author to create further text books. If you feel this content has helped you, please consider making a purchase.

Why write this book?

When I look around at colleagues I notice that we all come to app programming from different career paths. It seemed like everyone else found crash dump fixing burdensome apart from myself. I came to app programming from the bottom upwards. I started off as a device driver and kernel engineer and took subsequent jobs in the middleware layer and then lastly in the app layer. This less trodden route has given me a perspective which I would like to share with other developers and engineers.

It's already documented right?

  1. There are lots of documents on the specifics of exploring crash dumps, but they exist in a knowledge bubble not connected to the normal app developer experience; they feel alien. The knowledge is generally scattered around without a definitive guide. I would like collect the information together into one place.

  2. Software engineering concepts can be employed to avoid crashes, and to lessen the pain in finding and resolving them. I would like shine a light on this aspect of engineering so you can improve your codebase proactively.

  3. Step-by-step guides often seem to make large logical jumps to derive the answer to why an app crashed. But crash dump analysis is just problem-solving in a specific domain. There are systematic problem solving approaches that can be applied here. I have found these helpful in my own work and want to share the insights more widely.

It's my opinion that when you put these three approaches together: software engineering, systematic problem solving, and crash dump exploration you end up with a powerful toolbox you can take with you to any iOS app project. You will be able to spend less time fixing crashes, and more time writing valuable functionality. That's something we all want!

Project Details

Current Status

As of 14th November 2020, the project status is:

  • With the public announcement of Apple's new M1 chip and range of Macs, we can now release the second edition materials describing the new technologies!
  • The 2nd edition is published here in web form to cover Apple Silicon, etc.
  • There is a published 1st edition (print, kindle, web)
  • A web-only Chinese language translation is available (first edition elements are chinese, second edition elements not yet translated)
  • I am preparing the kindle and print editions of the second edition.

Directory Structure

Directory Purpose
external Downloaded resources
topics Topic specific writing
source Parent directory of compilable source code
markdown Markdown files for the body text of the book
examples Examples of crashes
automation Not yet used
screenshots Screenshots

Branch Policy

The software configuration management plan (SCM plan) is very simple.

  • All development is done on the master branch.
  • Feature work is done on a branch.
  • Notable versions of the software are tracked by Release History

Authoring Methodology

I followed an approach similar approach to http://rtalbert.org/how-i-wrote-my-book/ and Ben Watson at Philosophical Geek with a few amendments. The process turned out to be:

  • Write the content of the book in markdown
  • Use pandoc to convert the markdown into MS Word docx format, .epub, .latex, and .html
  • Use BibDesk to track biliographic references
  • Use latex \index to create index links in the main body of the work
  • Use pdflatex to convert latex files into a PDF because it allows auto section numbering and index generation

The output file formats were used as follows:

  • .docx version to do grammar and spell checking
  • .pdf version used to prepare the print version of the book via Kindle KDP web site
  • .epub version used to create the Kindle E-book via the Kindle KDP web substitute
  • .html version used to create the GitHub pages version

The writing was done on a per chapter basis, with separate markdown files for examples and their discussion. This allowed me to merge and move around the smaller level items into parent chapters to form a logical grouping.

Each chapter was done in phases:

  • Write the content, marking obvious index items, but generally focussing on content
  • Do needed digressions for research, and write sample code to justify statements made
  • Review the .docx version and run the spell and grammar checker, fixing any faults in the markdown
  • Review the markdown and insert latex index entries, cross referencing prior indexed items for uniformity and grouping

The aim was to leave each chapter in a near-publishable state, to avoid accumulating a large backlog of work. I resisted placing any TODO work items in the markdown text. I maintained a ideas.txt file for chapter and content ideas as I went through. I also did a grep for forward topic references in the book so that I covered later on what was promised earlier on. The main benefit was I had a good idea of my progress on the book as I went along.

In terms of writing style, I used first person singular for the preface. I used second person plural for the Introduction, and used first person plural elsewhere. I had a script to check not to use "you" when a "we" should have been used in the main body of the text.

Once all chapters were done, the whole book was read through using VoiceOver (Option-Escape) because this catches proof reading errors (like swapping around words or duplicating words).

I also hand wrote the back page text directly in .docx and then pruned it to fit the back of the book and the marketing materials.

I had a special index mark for trademark terms and a custom script to collate those and inject them into the text as a list of acknowledged trademarks. The entire book building process was automated from a command line script.

I had a special .css file for the E-book to make the tables look nicer (a simple outline border). I also used a github styled .css file for the HTML edition.

The last phase was upload to the KDP website to generate the e-book and print versions. I had to re-size my example code blocks to fit the Extended Distribution in Print requirements (6 inches by 9 inches is the biggest) as only standard book sizes are distributable via Extended Distribution (regular book shops).

Progress and Effort

The hardest part was learning about the appropriate tooling to do the job and deploy for the relevant target formats, and working out what formats and platforms to use. This consumed two weeks.

The tail portion of the work was quick. It took 1 day to re-base my code on the newly released Xcode 10 Gold Master. I just needed to re-work one chapter due to a behavior change in iOS 12. It took 1 day to proof read my work using VoiceOver. It took 1 day to re-format the code and report examples to change to 6 x 9 inch book format needed for Extended Distribution of the text in print format to physical book sellers, as well as the time to fix up the print ready PDF since Amazon has stricter gutter requirements (going into the margin) than I had anticipated. I also had to do the book cover text. I was overall pleased it was 3 days of final processing.

The chapters of the book were the most effort. Each chapter took about 0.5 days to fix spelling and grammar, and add index entries. The rest of the time was spent roughly 60% research, 40% writing but I am not sure as I didn't pay so much attention to the balance.

I wrote over a period of 3 months but had other tasks and duties during that time. So it was 7 weeks of full time work spread across 3 months. I worked normal business hours, roughly 40 hours per week. For 170 pages of text, it works out at 1.6 hours per page. If I were to write a second book, using the same technology and approach, I think it would be 5 weeks effort, and 1.2 hours per page. This is because in my first book I had to do 2 weeks research, and setup, of my build system to generate the book in the appropriate formats.

I planned to write more but I had some work opportunities that came up that were potentially going to limit my time to spend on the book and wasn't sure if it was going to pass the "moonlighting" test for private work. I would have written chapters explaining about Operating Systems architecture, assembly programming basics, and performance analysis with Instruments.

Second edition

I wrote the second edition of the book, full-time, over 1 month. I didn't track my time as closely but I think I was more efficient at doing things.

I improved my tooling, to automatically clean the gutter (by wrapping lines which were in markdown verbose sections). That took one day. I also cleaned up my scripting a bit. I added scripting to allow a translated edition to be produced in a way that would safely fall back to English for missing translations. I could not fix the Chinese latex issues. I think my lack of Chinese was going to make it unfeasible to resolve. I also spent one day learning how to use inkscape to cut out my portrait and then actually doing the cut out work, so I could insert it into the back cover. So I think I had total 4 days of non-writing overhead in the month.

I finished writing too early because Apple Silicon was released later than my expectations. In fact, the surprises were mostly on the content side. I knew I had to write about Apple Silicon but did not realise that Pointer Authentication was a relevant topic. I also missed that I could actually test iOS on Mac with my development setup - I thought that was going to come in later tooling. I also realised I needed a chapter to cover diagnostic related crashes (such as resource usage related terminations). So my writing plans did change, but it showed up some very interesting topics. It also allowed me to provide original research material in the book.

My chinese translator was a welcome surprise and integrating his work was very easy so I got "bonus" material by way of a 1st edition translation in Chinese.

The book is now a nice length, 239 sheets of paper, with the last page numbered 221. The first edition had a last page numbered 170.

The new material is 3.4 hrs per page. But there is original research in there so that low number feels like great productivity. Also some of the pages have been re-worked, with code modernisation, new screenshots, etc. These don't show in the numbers.

Overall I was surprised how many additions and modernisations I needed to do for the book. Especially with the archicture transition to Apple Silicon, I think this is a especially fast-moving period in the Apple developer ecosystem.

Producing the actual print edition, and kindle edition was easy because my approach, and tooling, made the book more-or-less fit for publishing straight away. I did not need to do any fix-ups.

I am making the second edition, in ebook, and in print, as cheap as I reasonably can, with only a small margin over costs. I think I am providing good value!

I think a two year update cycle feels right for this kind of book. Lets see what the future holds!

Supporting software

Essential Software

You need:

  • Brew Package manager
  • Various Brew packages
  • Latex packages
  • Class dump

Brew Software

I used the Brew package manager on MacOS and used the Brew packages:

Brew Package Purpose
pandoc Document translator to get from .markdown format to other formats
pandoc-citeproc Biliography Citation Helper for pandoc

Latex Support

Latex support appears not to be available directly in brew so I used the recommended MacTex and this was installed via a brew cask install brew cask install mactex

Class Dump

For analysing binaries I used class-dump. Whilst previously this was available from Brew, it seems now you have to directly download it.

Download class-dump

Recommended Software

Package Purpose
Atom Edits and understands markdown and can preview render it
BibDesk Eases the definition of Citations for the Biliography

Essential Mac Configuration

Screenshots

I have discovered the standard screenshot on Mac only does low resolution (screen resolution) images. Print books need to have higher resolution.

I saw an answer at: https://apple.stackexchange.com/a/110206

I need to experiment in this area. Maybe putting my screen in large scale mode will give me screenshots with effectively a higher ppi when shrunk down for print.

Build System

The book is built for English using buildBook.sh. Use buildBook.sh -l zh to build the Chinese edition.

The build outputs are:

File Purpose
foo.html Intermediate used for GitHub Pages documentation
boo.pdf For the Hard Copy Paper Edition
foo.docx For further conversion into EPUB, and internally for spelling and grammar checking
foo.epub For E-book readers (Apple and Amazon) directly from pandoc
docs/* Final destination for GitHub Pages documentation

The output are foo.* and boo.* files locally for ease of inspection. They are ignored by version control.

For github pages, the GitHub documentation facility, the HTML documentation and supporting resources is copied into the docs/<language-id> directory and then they are checked in (the branch is required to be master).

The build system also checks for uses of "you" - they should be "we" in all cases apart from the Introduction.

The build system only picks up new trademarks on a second run of the book building script due to it relying on a built index from the prior run.

The build system has one hardcode; the page number that the Disclaimer chapter starts on is hardcoded because it is based upon the number of pages used up by the automatically generated table of contents.