Re: [LTP] Genhtml patches

[Peter O. <obe...@go...>]

Hello,

Ludmil Alexandrov wrote:
> Here are two patches for genhtml.

Very nice! These are quite interesting new functions. See below for some more detailed comments.

> The first one
> (genhtml_1.5_branching.patch) adds branching highlighting in the html
> files for all the single source files (for example boot.c).

This looks good, though I don't quite understand the output :) I'll have to take a closer look before commenting on this one.


> The second
> one (genhtml_1.5_sidebyside.patch) compares visually side by side two
> info files.

Looks good and implements something I have thought about for some time as well. Some details that may be improved:
- the change in genhtml's syntax is quite big. Where one could specify a number of .info files for visualization, one now has to specify exactly two, so there's not even an option not to use this feature.

Suggestion: make it a named option like '--compare <tracefile>'. This option may be specified more than once. Like with the files specified directly, genhtml should read all the files and accumulate coverage information found in it. Then the comparison could be done between the sum of both sets of files.

- source view displays each line of code twice. This can result in rather long output lines. I think this can be optimized by using the following per line setup:

line number - coverage count of file 1 (background color for coverage 1) - coverage count of file 2 (background color for coverage 2) - source code line (white background)


What do you think?


> Let me know what you think. In addition, I noticed that lcov
> is quite slow, what are your thoughts for re-writing l-cov in C.

I think perl is better for this particular job because lcov is based a lot on text-file processing and pattern matching. Implementing this functionality in c would introduce quite a lot of code complexity which would make it harder to maintain and extend the code.

I have to admit though that the perl code itself is not optimized for speed at all. If anyone has tips on how to modify perl code to improve performance, feel free to share them.


Regards,
  Peter Oberparleiter