Latexian- difference live preview and pdf export in error reporting?

Question

I edit lecture slides for work and while the editor in the office runs on Linux (Kile), for home office I use Latexian.

I noticed that I can get perfect live previews in Latexian without any (noticeable) errors being reported (except when I go to the console), i.e. the document compiles.

However, as soon as I try to export the pdf, it doesn't compile and multiple errors are reported. Not only is this annoying but I am often at loss about what is wrong, given that when I did that faulty slide specifically, nothing was reported.

This makes me wonder if the live preview has some sort of built-in 'force' command to show the preview no matter what?

This is important for me as I fear that my supervisor might not be able to compile files on her Kile when I send him my work and think that I wasn't thorough enough checking for errors....

Example:

\pagenumbering{arabic}  
\documentclass[sans,mathserif,10pt]{beamer}  
\mode<presentation>  
\usepackage{graphicx}  
\usepackage{mdwlist}  
\usepackage{color}  
%\usepackage{ulsy}  
\usepackage{MnSymbol}  
\usepackage[T1]{fontenc}  
\usepackage{ragged2e}  
\usepackage{nicefrac}  
\input{preamble_js}  
\setbeamersize{text margin left=22pt,text margin right=22pt}  

\beamertemplatetransparentcovereddynamic
\title[ABC]{{ABC}}

\begin{document}  
\begin{frame}  
\titlepage
%\hyperlink{retpred}{\beamergotobutton{Skip}}
\end{frame}

\input{chapter4a.tex}

A typical input in the chapter4a- file would then be e.g.

\section{Is this MA($\infty$) process stationary and ergodic?}

\begin{frame}  
\vspace{0.6cm}  
Checking stationarity and ergodicity if |\phi|<1\\[0.5cm]  
\begin{equation*}  
\overset{\infty}{\underset{j=0}{\sum}}|\psi_j| =  
\overset{\infty}{\underset{j=0}{\sum}}|\phi^j| = \frac{1}{1-|\phi|}  
< \infty \quad \text{if} \quad |\phi|<1  
\end{equation*}  
$\Rightarrow$ stationary and ergodic! \\\\\\  
\vspace{0.5cm} The variance is given by:\vspace{0.1cm}  

\[\begin{aligned}  
\gamma_0 &=& \mathbb{E} \left[(Y_t - \mu)^2\right] = \mathbb{E} \left[(\varepsilon_t +\phi\varepsilon_{t-1}+\phi^2\varepsilon_{t-2}+\ldots)^2\right] \\  
&=&(1+\phi^2+\phi^4+\phi^6+\ldots)\sigma^2\\  
&=&\frac{1}{1-\phi^2} \; \sigma^2 \qquad (\text{if} \; |\phi|<1)  
\end{aligned}\]  
\end{frame}  

\end{document}

Answer 1

Just because LaTeX produces a reasonable-looking PDF doesn't mean it has compiled correctly. For example, the following document contains an error:

\documentclass{article}
\begin{document}

Subscripts like x_i need to be in math mode.

\end{document}

But if you run the command pdflatex -interaction nonstopmode test_document.tex, which I assume LaTeXian does, it will do its best to give you a readable PDF:

It's even possible to write flawed documents for which LaTeX will issue several compilation errors but still produce a PDF which looks just fine:

\documentclass{article}
\begin{document}

The source of $x}$_i$ is totally wrong but the output still looks okay.

\end{document}

The moral of the story is to always rely on the console, not the outputted PDF, to see whether your source has compiled correctly.