Hello, you're right about eval and symbol-value, of course. I mixed up my test files, so both the commit I `identified' and the `patch' are bogus. I now used git bisect for both my current minimal setup (.emacs attached as .femacs and the files ~/ot/1.org, ~/ot/2.org, and ~/ot/3.org all three having the contents of the attached file 1.org) and the setup and project I first encountered the issue. The issue occurs always during publishing the _second_ file, that is, the first file being published gets published Ok, and the following files don't. For the minimal example I cd'd to ~/ot and called emacs using the command line: rm *.tex ~/.org-timestamps/linux-e-pdf.cache ; emacs -q -l ~/.femacs For the original project I used my full setup but also deleted *.tex, *.pdf, and ~/.org-timestamps/linux-e-pdf.cache. To be sure nothing went wrong I bisected twice for each project. For the minimal example commit 2965f8fb0c048a20b52ba90627e7cca6fe706c93 is identified as the first bad one. For my full setup it is the earlier commit 69dd4301ab7aeb581968f1c7e75167338d024c8f. My Emacs version is : GNU Emacs 24.5.1 (x86_64-suse-linux-gnu, GTK+ Version 3.10.2) of 2015-04-30 on build83 Meanwhile I added a second elisp source and call (between latex_class_options and the later options line). With this added I don't see the error anymore. Same for the minimal example (org files all having the content of the attached file working.org) I'm sorry for the earlier confusion. Best regards Robert On Sun, 21 Jun 2015 00:15:49 +0200 Nicolas Goaziou wrote: > Hello, > > Robert Klein writes: > > > when I change a `symbol-value' in the commit back to `eval', export > > seems to work again. > > Actually, I cannot reproduce your issue. > > > Patch is attached. > > I don't understand why replacing `symbol-value' with `eval' could > solve the issue. Could you debug this and see what happens with both > `symbol-value' and `eval'? > > Thank you. > > Regards, >