rtags: "etags: no input files specified."

Question

When I do rtags(ofile="TAGS") at the R prompt, the "TAGS" file is written and there is no output to the terminal (exactly as expected).

When I do R CMD rtags -o TAGS at the shell prompt, the "TAGS" file is written too, but I see several sets of messages on the terminal like this:

etags: no input files specified.
    Try `etags --help' for a complete list of options.

I see 6 sets - 12 lines - when I move my libPath out of the current directory and two sets - 4 lines - when I keep it there. I.e., I see more warnings when rtags processes fewer files.

To reproduce, run in an empty directory:

$ mkdir z
$ cd z
$ R --vanilla CMD rtags

Tagging R/C/Rd files under /home/sds/z; writing to TAGS (overwriting)...

etags: no input files specified.
    Try `etags --help' for a complete list of options.
etags: no input files specified.
    Try `etags --help' for a complete list of options.
etags: no input files specified.
    Try `etags --help' for a complete list of options.
etags: no input files specified.
    Try `etags --help' for a complete list of options.
etags: no input files specified.
    Try `etags --help' for a complete list of options.
etags: no input files specified.
    Try `etags --help' for a complete list of options.
Done

What causes these warnings? Is there a way to avoid them?

Answer 1

The command R CMD rtags create a TAGS file for 3 different types of files: R files, C files and Rd files (documentation for R files). For R files it uses the R function utils::rtags() and for C and Rd files it calls the etags utility (on Linux; not sure what it does on MacOS or Windows).

The error messages that you see are emitted by etags when it is called without any input file. This happens because R CMD rtags uses find to look for files to process by etags and feeds the output of find into etags. If there are no C or Rd files in the directory you are processing or in any sub-directory of it, etags would be called with an empty list of files to process, which will cause it to print an error message.

You see several error messages because there are separate calls to etags for Rd files and 'C files', which actually inclueds .c, .h, .cpp and several other types.

In order to suppress these messages you should explicitly tell R CMD rtags not to process files you don't have. For example, R CMD rtags --no-c will not try to look for C files, and R CMD rtags --no-Rd will not try to look for Rd files.

So if, for example, you are interested in tagging only R files, use R CMD rtags --no-c --no-Rd. See R CMD rtags --help for more details.

As a side note, if you have C files in your project and you do want to tag them along with R files, you might still get such error messages - say that you have *.c and *.h files but no *.cpp files, and you call R CMD rtags without the --no-c flag. The command will also look for *.cpp files and call etags with an empty list.

Another note that worth mentioning here, although not directly related to the question, is that R CMD rtags look only for *.R files that are directly under a directory named R. So R files in a directory with a different name would not be tagged. If you need to tag R files that are in such directories, or if you need more flexibility, you can call utils::rtags() from an R session with suitable arguments.

Answer 2

When I run this from a console session the warnings to the console are not at all like yours, but they are mostly basically meaningless comments about the process of walking through the files in my working directory:

1: In file.remove(ofile) :
cannot remove file 'TAGS', reason 'No such file or directory'
.....
: In readLines(file) :, incomplete final line found on './Untitled.R' 

I did have few like this:

6: In grepl("\n", lines, fixed = TRUE) : input string 5 is invalid in this locale

The real information about the location of assigned tokens in the code goes into the TAGS file. Since the warnings in my setup are quite different, I still think the answer to your question about the increase in the number of warnings when you change the .Library variable will depend on the specific code that R is parsing during the operation. A guess: removing code from being loaded may be making certain operations not possible that would otherwise have run smoothly. And remember these are only 'warnings'.