rtags-install fails - cmake: command not found

[sfavazza]

Describe the bug
When running the emacs command to install the rtags package I get the error cmake: command not found, even if the CMake tool is correctly installed.

To Reproduce
Steps to reproduce the behavior:

1. open emacs
2. M-x load-library RET rtags
3. M-x rtags-install RET
4. In the *RTags Install* buffer you should see a line like the following:

   /home/sfavazza/.emacs.d/elpa/rtags-20200221.36//install-rtags.sh: line 21: cmake: command not found
   Failed to cmake

Expected behavior
I would expect the installation going smoothly without any interruptions.

Observations and Workarounds
Despite the confusing error (as CMake is correctly installed on my machine), I realized that the automatically generated *.sh installation script is not marked as executable.
Moreover the attempt to repeat the same script ends up into another error:

curl: (33) HTTP server doesn't seem to support byte ranges. Cannot resume.

This instead is due to the fact that the archive was already downloaded once already.
Normally I would expect that the installation script tries to delete such file before attempting a download.

Based on these observations I managed to find a workaround (I guess it can be easily solved by modifying the rtags-install function).

From emacs:

1. M-x load-library RET rtags
2. M-x rtags-install (this will fail, but it is required to create the *.sh file)

From shell
3. cd ~/<rtags-installation-path>/
4. chmod +x install-rtags.sh
5. rm rtags-<version>.tar.bz2
6. ./install-rtags.sh

If anyone else had a similar issue, I might create a fix, I just would like to know which tests should I run to ensure nothing got broken by my fix.

Environment (please complete the following information):

• OS/Version: Debian 9 (stretch)
• RTags(rdm,rc): 2.38.130
• Emacs: 26.2
• LLVM/Clang: 3.8.1-24

[zhihuiFan]

what is the location of your cmake? maybe your emacs uses a different search path for binary.

you may try add the following code to your init.el

(require 'exec-path-from-shell)

(after-load 'exec-path-from-shell
  (dolist (var '("SSH_AUTH_SOCK" "SSH_AGENT_PID" "GPG_AGENT_INFO" "LANG" "LC_CTYPE"))
    (add-to-list 'exec-path-from-shell-variables var)))


(when (memq window-system '(mac ns x))
  (exec-path-from-shell-initialize))

[sfavazza]

Yeah, I realized that my exec-path was not properly set because I start emacs in server-mode so it inherit only the PATH system variable before the user configuration is loaded.

Yet, I don't know whether all the observations I mentioned in the issue description are due to the failed Makefile execution.

[zhihuiFan]

I think we can fix the "cmake command not found" issue first by manually set exec-path at least , and then try if we still have another problems.

[sfavazza]

I ran the command again, from a fresh emacs instance and everything went smooth.

But I had to delete the curl-downloaded file first. Otherwise I was getting the confusing error message I mentioned in this thread.

Perhaps the Makefile could be modified with a mechanism to test whether the file to be downloaded is already there. If there ask the user to delete it before proceeding any further.

[sfavazza]

Closing for inactivity