[gcc] Gcc error: gcc: error trying to exec 'cc1': execvp: No such file or directory

I have been successfully using gcc on Linux Mint 12. Now I am getting an error. I have recently been doing some .so builds and installed Clang not to long ago, but have successfully compiled since both of those events, so not sure what has changed. I used the GUI Software Manager to remove and then install gcc again, but the results are the same:

~/code/c/ut: which gcc                                                                                                     
/usr/bin/gcc

~/code/c/ut: gcc -std=c99 -Wall -Wextra -g -c object.c                                                                      
gcc: error trying to exec 'cc1': execvp: No such file or directory

This question is related to gcc

The answer is


Just to document my trouble with this issue even though it just appears to be a specific example of other answers; as a relative newbie I feel like this might help others.

Solution:

I added '/usr/bin' to the beginning of PATH for a single session using PATH='/usr/path/:$PATH' and everything started to work fine.

I used gedit to update the PATH permanently, after ensuring it wouldn't break my regular toolchains.

Explanation:

I have multiple toolchains installed on Ubuntu 14.04LTS and I use just a couple on a regular basis. When I tried to use gcc from the command line I got the issue describe by the OP. '/usr/bin' is in the PATH but it is behind the other toolchain locations. Turns out the cc1 for those other toolchains is incompatible with gcc.


You can fix that by running this: On Fedora:

sudo dnf install redhat-rpm-config

Make sure your GCC_EXEC_PREFIX(env) is not exported and your PATH is exported to right tool chain.


This might also be the displayed error message if you try to run 32-bit gcc binaries on a 64-bit OS and missing 32-bit glibc. According to this readme: "For 64 bit system, 32 bit libc and libncurses are required to run the tools.". In this case there is no problem with the path and cc1 is actually found, but reported as missing as no 32 bit glibc.


This is because gcc calls many other executables to complete the processing of the input, and cc1 is not in the included path.

On shell type whereis cc1. If cc1 is found, it's better go ahead and create a softlink in the directory of gcc; otherwise, cc1 is not installed and you have to install gcc-c++ using the package manager.


Why does this happen? When you install a fresh copy of linux, gcc compiler comes pre-packed with it. It only contains the files and binaries which are used to run the linux(to save space and time, obviously).

How to solve this error? All you need is to update your packages through the package manager and reinstall the build-essential packages. The commands might be different on different kernels.


On Scientific Linux 6 (similar to CentOS 6-- SL is now replaced by CentOS, AIUI), I had to use /usr/sbin/prelink -av -mR which I found suggested at https://stelfox.net/blog/2014/08/dependency-prelink-issues/

Until I did that, I got a cc1 error gcc: error trying to exec 'cc1': execvp: No such file or directory when I tried to compile, and gcc --version reported 4.2.2 instead of 4.4.7, despite that version being reported by yum.

It may or may not be related, but the system had run out of space on /var


On CentOS or Fedora

yum install gcc-c++ 

What helped for me was to use llvm-gcc instead:

ln -s $(which llvm-gcc) /usr/local/bin/gcc

I experienced this soon after compiling and installing a shiny new GCC — version 8.1 — on RHEL 7. In the end, it ended up being a permissions problem; my root umask was the culprit. I eventually found cc1 hiding in /usr/local/libexec:

[root@nacelle gdb-8.1]# ls -l /usr/local/libexec/gcc/x86_64-pc-linux-gnu/8.1.0/ | grep cc1
-rwxr-xr-x 1 root root 196481344 Jul  2 13:53 cc1

However, the permissions on the directories leading there didn't allow my standard user account:

[root@nacelle gdb-8.1]# ls -l /usr/local/libexec/
total 4
drwxr-x--- 3 root root 4096 Jul  2 13:53 gcc
[root@nacelle gdb-8.1]# ls -l /usr/local/libexec/gcc/
total 4
drwxr-x--- 3 root root 4096 Jul  2 13:53 x86_64-pc-linux-gnu
[root@nacelle gdb-8.1]# ls -l /usr/local/libexec/gcc/x86_64-pc-linux-gnu/
total 4
drwxr-x--- 4 root root 4096 Jul  2 13:53 8.1.0

A quick recursive chmod to add world read/execute permissions fixed it right up:

[root@nacelle 8.1.0]# cd /usr/local/libexec
[root@nacelle lib]# ls -l | grep gcc
drwxr-x---  3 root root     4096 Jul  2 13:53 gcc
[root@nacelle lib]# chmod -R o+rx gcc
[root@nacelle lib]# ls -l | grep gcc
drwxr-xr-x  3 root root     4096 Jul  2 13:53 gcc

And now gcc can find cc1 when I ask it to compile something!


I experienced this problem on a reasonably fresh install of Fedora 27. I tried all the other suggestions or their equivalents; installing the various packages either said "already installed" or installed something new which didn't help.

Fixed with

# dnf remove gcc
# dnf install gcc gcc-c++

In my rare case it was color wrapper who spoiled gcc. Solved by disabling cw excluding its directory /usr/libexec/cw from PATH environmental variable.


Explanation

The error message told us, that the build-time dependency (in this case it is cc1) was not found, so all we need — install the appropriate package to the system (using package manager // from sources // another way)

What is cc1:

cc1 is the internal command which takes preprocessed C-language files and converts them to assembly. It's the actual part that compiles C. For C++, there's cc1plus, and other internal commands for different languages.

taken from this answer by Alan Shutko.

Solution for: Ubuntu / Linux Mint

sudo apt-get update
sudo apt-get install --reinstall build-essential

Solution for: Docker-alpine environment

If you are in docker-alpine environment install the build-base package by adding this to your Dockerfile:

RUN apk add build-base

Better package name provided by Pablo Castellano. More details here.

If you need more packages for building purposes, consider adding of the alpine-sdk package:

RUN apk add alpine-sdk

Taken from github

Solution for: CentOS/Fedora

This answer contains instructions for CentOS and Fedora Linux

Solution for: Amazon Linux

sudo yum install gcc72-c++

Taken from this comment by CoderChris

You could also try to install missed dependencies by this (though, it is said to not to solve the issue):

sudo yum install gcc-c++.noarch

Taken from this answer


Amazon Linux: fixing GCC issue

Since this comes up as the first result on Google, I just wanted to document my experience with Amazon Linux. Installing gcc-c++.noarch fixed the problem:

sudo yum install gcc-c++.noarch

Some people also reported this alternative as a solution:

sudo yum install gcc72-c++


yum install gcc-c++ did the fix.


It's in this package (Ubuntu 19.04):

  sudo apt install g++-6

I fixed this problem by explicitly installing g++:

sudo apt-get install g++

Problem was encountered on Ubuntu 12.04 while installing pandas. (Thanks perilbrain.)


Documenting another source of errors for installing gcc-10 on Amazon Linux 2 from source.

After running sudo make install and then testing gcc-10 I got this error:

gcc-10: fatal error: cannot execute ‘cc1’: execvp: No such file or directory

The reason was that the new g++ directories under /usr/local/ were created by sudo make install have 700 permissions so non-root users cannot see the directories content.

I fixed it by running

sudo find /usr/local/ -type d -exec chmod 755 {} \;

Note that I followed this snippet https://gist.github.com/nchaigne/ad06bc867f911a3c0d32939f1e930a11


On debian / ubuntu I fixed this problem by reinstalling build-essential:

sudo apt-get update
sudo apt-get install --reinstall build-essential

I ran into a similar issue today - a co-worker could not build his software but I could build it. When he ran gcc it could not find cc1.

His executable path looked reasonable but the fact that I could not easily replicate the failure suggested something in his environment as the cause.

Eventually we found GCC_EXEC_PREFIX defined in his environment which was the culprit and was misleading gcc in the search for cc1. This was part of his shell startup scripts and was meant to work around a limitation on a SPARC/Solaris system that is no longer in use. The problem was resolved by not setting this environment variable.

http://gcc.gnu.org/onlinedocs/gcc/Environment-Variables.html


Just to complement @maxkoryukov's answer regarding Alpine.

The equivalent to Debian's build-essential in Alpine is build-base. In fact, the above mentioned alpine-sdk depends on build-base.

/ # apk info -R build-base
build-base-0.5-r1 depends on:
binutils
file
gcc
g++
make
libc-dev
fortify-headers

/ # apk info -R alpine-sdk
alpine-sdk-1.0-r0 depends on:
abuild
build-base
git