From 7fe6ca2ca6d0bff32f0ab47606ac0511465f73d3 Mon Sep 17 00:00:00 2001
From: Mirek Kratochvil <exa.exa@gmail.com>
Date: Sun, 10 May 2020 22:54:11 +0200
Subject: [PATCH] PDF/A additions

---
 README.md | 17 ++++++++++++++++-
 pdfa.sh   |  4 ++++
 2 files changed, 20 insertions(+), 1 deletion(-)

diff --git a/README.md b/README.md
index cac4172..4b8606c 100644
--- a/README.md
+++ b/README.md
@@ -18,7 +18,22 @@ See the [pre-built version](build/thesis.pdf) for details
 4. Write the thesis.
 5. Submit and defend the thesis.
 
-PDF-A validation can be checked using the validator: https://github.com/mff-cuni-cz/cuni-thesis-validator
+## PDF/A
+
+With a bit of luck, you should get PDF/A right out of LaTeX.
+
+A working PDF/A validator that can point out exact problems is here: https://github.com/mff-cuni-cz/cuni-thesis-validator
+
+Common PDF/A problems include:
+
+- imported PDF pictures that are not PDF/A.
+- the used font does not support PDF/A (including the fonts in imported pictures). See https://martin.hoppenheit.info/blog/2018/pdfa-validation-and-inconsistent-glyph-width-information/ for a very ugly case.
+
+Solutions:
+
+- use `pdfa.sh` to convert PDFs to PDF/A-compatible form the "hard way" (although this does _not_ retain the PDF/A metadata mark, see comments in the script).
+- read the commentary by Martin Mareš (that describes most of the common problems) here: https://mj.ucw.cz/vyuka/bc/pdfaq.html
+- use `pdfa.sh` as a last resort for `thesis.pdf` if everything other fails
 
 ## Ideas/improvements/more examples?
 
diff --git a/pdfa.sh b/pdfa.sh
index 1fa5934..dce617a 100755
--- a/pdfa.sh
+++ b/pdfa.sh
@@ -13,3 +13,7 @@ gs -dPDFA=1 \
    -dPDFACompatibilityPolicy=3 \
    -sOutputFile="pdfa-$1" \
    "$1"
+
+# note: PDFACompatibilityPolicy=3 actually doesn't exist. A bug in ghostscript
+# interprets is as something between 1 and 2, without unnecessary failing on
+# various dumb errors.