iTerm2 shell integration breaks android builds

1. Install iterm2 integration
2. Try to do the "lunch" step at http://source.android.com/source/building-running.html

The following error is seen

build/core/config.mk:60: /pathmap.mk: No such file or directory
build/core/config.mk:151: /envsetup.mk: No such file or directory
build/core/config.mk:215: *** No TARGET_CPU_ABI defined by board config: .  Stop.

Removing iTerm2 integration from the bash_profile fixes the issue.

Milestone changed to Ship 3.0

Still seen on the latest nightlies - I tried to debug and I believe the preexec may be interfering with the android envsetup but I couldn't get any further

Sounds possible. OMZ ran in to a similar issue with a chpwd hook issuing terminal control commands that were screwing up Android build tools by contaminating captured shell function output.

https://github.com/robbyrussell/oh-my-zsh/issues/3524

A good approach might be to defensively use [ -t 1 ] to check whether stdout is actually attached to a tty before outputting the control sequences. Or write them directly to /dev/tty instead of to standard out.

Sorry I haven't gotten a chance to look into this more--it's super time consuming to get the environment set up to repro this bug so other things keep taking priority. Does does @apjanke 's trick work? Modify your .bashrc to say

test -t 1 && source ~/.iterm2_shell_integration.bash

No that didn't work - probably because the iterm2 shell integration was sourced first and then I tried to source the AOSP build environment.

From what I recall, some AOSP stuff is implemented as functions to be run inside an interactive shell, not in subshells. So the [ -t 1 ] guard would need to be pushed down in to the integration functions which are producing the output that AOSP doesn't like.

Naseer: http://source.android.com/source/building-running.html is 404 for me now. Is this transient? Or is there a new location for the Android build instructions?

I added [ -t 1 ] to all of the functions in iterm2_shell_integration.bash and it still didn't work - I may have done something wrong though.

The source.android.com page seems to have moved to http://source.android.com/source/building.html

That new link is working for me. I'll have a look and see if I can repro when I get a chance.

Naseer, could you post the new modifications you made, maybe as a branch on a GitHub fork or a gist?

Sorry - I was on vacation after my last post and don't have the change any more - will try it out again and update.

@apjanke The diff below was what I tried

diff --git a/iterm2_shell_integration.bash b/iterm2_shell_integration.bash
index 1e72696..75c88c9 100644
--- a/iterm2_shell_integration.bash
+++ b/iterm2_shell_integration.bash
@@ -45,6 +45,7 @@ if ( [ x"$TERM" != xscreen ] ); then
   # was just displayed, to allow the DEBUG trap, below, to know that the next
   # command is likely interactive.
   function iterm2_preexec_invoke_cmd () {
+      [[ -t 1 ]] || return
       local s=$?
       last_hist_ent="$(history 1)";
       precmd;
@@ -158,6 +159,7 @@ if ( [ x"$TERM" != xscreen ] ); then

   # Execute this to set up preexec and precmd execution.
   function iterm2_preexec_install () {
+      [[ -t 1 ]] || return

       # *BOTH* of these options need to be set for the DEBUG trap to be invoked
       # in ( ) subshells.  This smells like a bug in bash to me.  The null stackederr
@@ -190,12 +192,14 @@ if ( [ x"$TERM" != xscreen ] ); then

   # Runs after interactively edited command but before execution
   function preexec() {
+    [[ -t 1 ]] || return
     iterm2_begin_osc
     printf "133;C"
     iterm2_end_osc
   }

   function iterm2_print_state_data() {
+    [[ -t 1 ]] || return
     iterm2_begin_osc
     printf "1337;RemoteHost=%s@%s" "$USER" "$iterm2_hostname"
     iterm2_end_osc
@@ -209,6 +213,7 @@ if ( [ x"$TERM" != xscreen ] ); then

   # Usage: iterm2_set_user_var key value
   function iterm2_set_user_var() {
+    [[ -t 1 ]] || return
     iterm2_begin_osc
     printf "1337;SetUserVar=%s=%s" "$1" $(printf "%s" "$2" | base64)
     iterm2_end_osc
@@ -221,6 +226,7 @@ if ( [ x"$TERM" != xscreen ] ); then
   }

   function iterm2_prompt_prefix() {
+    [[ -t 1 ]] || return
     iterm2_begin_osc
     printf "133;D;\$?"
     iterm2_end_osc
@@ -233,12 +239,14 @@ if ( [ x"$TERM" != xscreen ] ); then
   }

   function iterm2_prompt_suffix() {
+    [[ -t 1 ]] || return
     iterm2_begin_osc
     printf "133;B"
     iterm2_end_osc
   }

   function iterm2_print_version_number() {
+    [[ -t 1 ]] || return
     iterm2_begin_osc
     printf "1337;ShellIntegrationVersion=1"
     iterm2_end_osc

Why not condition the whole script on [ -t 1 ]?

Tried that too - it didn't work since I'm sourcing shell integration when starting the bash session, then trying to setup the android build environment. Once it has been sourced, it will no longer allow me to run "lunch" from the android build environment functions.

Is this patch sufficient?

index 9744690..e809437 100644
--- a/source/misc/bash_startup.in
+++ b/source/misc/bash_startup.in
@@ -191,6 +191,7 @@ if ( [ x"$TERM" != xscreen ] ); then

   # Runs after interactively edited command but before execution
   function preexec() {
+    [[ -t 1 ]] || return
     iterm2_begin_osc
     printf "133;C;\r"
     iterm2_end_osc
@@ -199,6 +200,7 @@ if ( [ x"$TERM" != xscreen ] ); then
   }

   function iterm2_print_state_data() {
+    [[ -t 1 ]] || return
     iterm2_begin_osc
     printf "1337;RemoteHost=%s@%s" "$USER" "$iterm2_hostname"
     iterm2_end_osc
@@ -212,6 +214,7 @@ if ( [ x"$TERM" != xscreen ] ); then

   # Usage: iterm2_set_user_var key value
   function iterm2_set_user_var() {
+    [[ -t 1 ]] || return
     iterm2_begin_osc
     printf "1337;SetUserVar=%s=%s" "$1" $(printf "%s" "$2" | base64)
     iterm2_end_osc
@@ -224,6 +227,7 @@ if ( [ x"$TERM" != xscreen ] ); then
   }

   function iterm2_prompt_prefix() {
+    [[ -t 1 ]] || return
     iterm2_begin_osc
     printf "133;D;\$?"
     iterm2_end_osc
@@ -236,6 +240,7 @@ if ( [ x"$TERM" != xscreen ] ); then
   }

   function iterm2_prompt_suffix() {
+    [[ -t 1 ]] || return
     iterm2_begin_osc
     printf "133;B"
     iterm2_end_osc

No - it didn't work :(

@naseer-ahmed What if you add back the one in iterm2_print_version_number? Is that enough?

No - the diff I had posted above didn't work either - I just posted what I tried :(

I have it down to a simple reproducible case, but I'm not sure how to fix it.

echo 'echo $BUILD_SYSTEM' > p
chmod +x p
# Note that this prints "build/core"
BUILD_SYSTEM=build/core command ./p

# Install a trap. This changes the environment that p sees.
trap 'george "$_"' DEBUG;
function george {
  true
}

BUILD_SYSTEM=build/core command ./p

I filed a bug against bash-preexec. Hopefully they can think of a workaround:

https://github.com/rcaloras/bash-preexec/issues/19

mentioned in issue #4267 (closed)

Might be worth mentioning that it is very isolated problem in Android build and until the bash-preexec problem is fixed, there is an easy workaround. The below patch in build of Android build fixes the issue iterm2_shell_integration.patch

I'm going to close this out because there doesn't seem to be a way to fix bash, but it is possible to work around the Android problem with this patch.

Status changed to closed