On 2018-01-22, Kornel Benko wrote: > I expanded the lyx2lyx test so, the an error is emitted at 10th > repeated run to export lyx16x.
> If two consecutive created lyx-files are identical, the test stops > without error. if a created lyx-file is not loadable, test stops with > error. We found and fixed some problems with the new test. However, there are still problems that go undetected: * We currently do not test, whether a backported file can be opened by the target LyX version. This means that instances of new constructs not handled by lyx2lyx go undetected. (An example are "baseskip%" lengths in box insets.) Is there a chance, to test the lyx2lyx output with LyX 2.2, 2.1 and maybe even 1.6? * LyX does accept some variability in the input files and normalizes them on output. Examples are newlines or the order of settings in the document header. - Do we want lyx2lyx to produce output that is unchanged when opened and saved by the target LyX version? What is the price we are willing to pay for this in terms of - added complexity of the lyx2lyx functions - ly2lyx run time Examples: * Reverting Quote insets to literal quotes adds newlines (or rather does not remove them). * The Developers guide says we should add new settings in lyx2lyx but this can also be taken care of by LyX itself: a missing setting is just set to the default value already. * Sorting the settings to the place where LyX will save them can be done in lyx2lyx but makes the conversion take longer and the lyx2lyx script have 2 more lines. It is not necessary, though (or am I missing something here)? * Sometimes, the back-converted file will load in LyX but fail to compile due to errors in the generated LaTeX. Could we also test this? Günter