[petsc-dev] broken petsc-dev related mesh?

[Barry Smith]

On Mar 12, 2011, at 1:32 PM, Jed Brown wrote:

> On Sat, Mar 12, 2011 at 20:21, Barry Smith <bsmith at mcs.anl.gov> wrote:
> I like them in place and think you need a more subtle tool then grep -noflags to find stuff :-).
> 
> Well, "ack" will ignore them by default, "ls" does not.

    ls with proper options will.  For example ls -sourceonly


>  
> Come on, our source management tools haven't improved since 1975?
> 
> Good source management tools are aware of whatever is currently being built and do not look at things that are disabled (e.g. files ignored by the build system or stuff inside of #ifdef NOT_DEFINED). It is also useful to search more than just source files, for example to include references to a symbol inside the user's manual or web page.

   Absolutely correct. 

> When you explicitly do not want semantic analysis, tools like grep are still handy. I don't know any smart tools that have a "semantic" mode and a "non-semantic, but when looking at html files, disregard those that have been generated, however please still examine generated fortran stubs".

   Absolutely correct and ETAGS/VTAGS/Eclipse stuff should be managed in such a way so that all these different search paradigms can be done using a single simple consistent way.  So, for example, I should be able to say search in source only, search in all non-generated files only, search in certain directories only, search in generated source code only etc etc. 

   The fact that we cannot do all these things easily and consistently is why we sometimes have to revert to falling back to grep. I was not attacking you for using grep but pointing out if our search system were better then you would not need grep and hence would not care of the .html is there. I don't want to take the .html somewhere else just because our search system stinks.

   Note that make alletags does now generate several tags files, yes they are not great categories and this needs tons of improvements but I'd like to see these done properly. So it won't be truly semantic and non-semantic modes but it would make all common search paradigms easy.


> 
> Why do you want them inside the source tree instead of with the rest of the generated documentation?

   The rest of the generated documentation is not actually stored in a different directory tree (if you use LOC=$PETSC_DIR) it is stored in the PETSC_DIR directory. I consider the entire PETSC_DIR (source code and all after make alldocs is built) to be the fully documented package. For example the generated .html has links at the top to the original code. If there is a bug in the handling of where this goes let us know so we can fix it.

   Barry