Avoid creating nested symlinks (at least by default) [sf#228]

[tomschr]

Reported by sknorr on 2014-07-01 10:33 UTC
We now have some symlinks in the static/ directories for our HTML output (to avoid duplicating too many files). In the current default setting, this means DAPS creates nested symlinks when building html books, e.g. with:

$ daps -d DC-example html

This can only be avoided by doing

$ daps -d DC-example html --static

Nested symlinks are not a problem per se, but they introduce various resolution issues in both Firefox and Chrome.

So, there would be two possible resolutions to this bug:

1. Always enable the behaviour of --static (and, if really really necessary add an option --nostatic).
2. When building the static directory, resolve the links in all files first, and then symlink the original files

Option 1 almost seems like the better option to me, as it avoids any ambiguities in references[1].

---

[1] as an example for an ambiguous reference: e.g.

• page.html loads style.css,
• style.css itself is a symlink to ../../otherfolder/real.css and
• style.css tries to load image.png

Question: Where should the browser search for image.png? in the current folder or in ../../otherfolder?

[tomschr]

Commented by sknorr on 2014-08-19 16:33

• Description has changed:

Diff:

--- old
+++ new
@@ -9,8 +9,9 @@
 Nested symlinks are not a problem per se, but they introduce various resolution issues in both [Firefox](https://bugzilla.mozilla.org/show_bug.cgi?id=symlink) and Chrome.

 So, there would be two possible resolutions to this bug:
-1) Always enable the behaviour of --static (and, if really really necessary add an option --nostatic).
-2) When building the static directory, resolve the links in all files first, and then symlink the original files
+
+1. Always enable the behaviour of --static (and, if really really necessary add an option --nostatic).
+1. When building the static directory, resolve the links in all files first, and then symlink the original files

 Option 1 almost seems like the better option to me, as it avoids any ambiguities in references[1].

• assigned_to: Frank Sundermeyer
• Info Provider: Stefan Knorr