I have a task to grab a frontend from git for my Play Framework application:
lazy val frontend = taskKey[Unit]("Downloads frontend")
frontend := {
val s: TaskStreams = streams.value
val shell: Seq[String] = if (sys.props("os.name").contains("Windows")) Seq("cmd", "/c") else Seq("bash", "-c")
val downloadRepo: Seq[String] = shell :+ "git clone git#bitbucket.org:user/frontend.git"
val rmJs: Seq[String] = shell :+ "rm -rf frontend/dist/js && rm -rf public/design && mkdir public/design"
val copy: Seq[String] = shell :+ "mv frontend/dist/* public/design/"
val rmRepo: Seq[String] = shell :+ "rm -rf frontend"
s.log.info("Downloading frontend...")
if((downloadRepo #&& rmJs #&& copy #&& rmRepo !) == 0) {
s.log.success("frontend downloaded successful!")
} else {
throw new IllegalStateException("frontend failed!")
}
}
and it works fine with sbt 0.13.x, but I want to migrate to newest one and it gives me an error:
error: value #&& is not a member of Seq[String]
if((downloadRepo #&& rmJs #&& copy #&& rmRepo !) == 0) {
I checked new documentation and didn't find an answer, how can I migrate this?
I guess you mean the #&& operator from sys.process:
#&& conditionally executes the second command if the previous one finished with exit value 0. It mirrors shell's &&.
sbt 0.13 had an API similar to that of sys.process and in 1.0 sys.process replaced it. So the fix would be just to import it to add relevant implicints to the scope:
import scala.sys.process._
Related
I'm using a Mac and followed the instructions to install with homebrew and added eval "$(starship init zsh)" to ~/.zshrc.
However, when I restarted my terminal it still shows the default one.
Here's the content of my .zshrc (added the line at the bottom)
# >>> conda initialize >>>
# !! Contents within this block are managed by 'conda init' !!
__conda_setup="$('/opt/anaconda3/bin/conda' 'shell.zsh' 'hook' 2> /dev/null)"
if [ $? -eq 0 ]; then
eval "$__conda_setup"
else
if [ -f "/opt/anaconda3/etc/profile.d/conda.sh" ]; then
. "/opt/anaconda3/etc/profile.d/conda.sh"
else
export PATH="/opt/anaconda3/bin:$PATH"
fi
fi
unset __conda_setup
# <<< conda initialize <<<
export PATH="$HOME/.rbenv/bin:$PATH"
eval "$(rbenv init -)"
export GOPATH=$HOME/go
export PATH=$GOPATH/bin:$PATH
export GOBIN=$GOPATH/bin
eval "$(starship init zsh)"
I add depends packet to a exist .bb file, such as add DPENDS="AAA" line to .bb file, when I compile the .bb file, it failed for XXX rdepends on AAA-dev [dev-deps], and I search google, all the answer almost is add line INSANE_SKIP_${PN} += "dev-deps" or RDEPENDS_${PN}_remove = "AAA-dev" to .bb file.
But my question is why? why one packet depend AAA packet, it should also RDPENDS AAA-dev, is there any other answer to fix this problem
The bb source file is:
inherit autotools qcommon
DESCRIPTION = "Daemon to handle AT commands"
DEPENDS = "glib-2.0 qmi qmi-framework qmi-client-helper ocean-link"
SRC_DIR = "${WORKSPACE}/atfwd-daemon/"
S = "${WORKDIR}/atfwd-daemon/"
PR = "r3"
EXTRA_OECONF += "--with-glib --with-common-includes=${STAGING_INCDIR}"
do_configure_append() {
echo "/*This is compiled to generate, only look don't try*/" > ${S}atfwd_config.h
echo "#ifndef _ATFWD_CONFIG_H_" >> ${S}atfwd_config.h
echo "#define _ATFWD_CONFIG_H_" >> ${S}atfwd_config.h
#//<!-- ODM feature caogang#2015-07-13
if [ "${PRJ_NAU8810}" = "NAU8810_CODEC" ]; then
echo "#define NAU8810_CODEC" >> ${S}atfwd_config.h
fi
if [ "${FEATURE_ACDB_ENABLE}" = "true" ]; then
echo "#define FEATURE_ACDB_ENABLE 1" >> ${S}atfwd_config.h
fi
if [ "${PRJ_XXX}" != "" ]; then
echo "#define ${PRJ_XXX}" >> ${S}atfwd_config.h
fi
#//end-->
I add a DEPENDS on onenet pkg
inherit autotools qcommon
DESCRIPTION = "Daemon to handle AT commands"
DEPENDS = "glib-2.0 qmi qmi-framework qmi-client-helper ocean-link onenet"
SRC_DIR = "${WORKSPACE}/atfwd-daemon/"
S = "${WORKDIR}/atfwd-daemon/"
PR = "r3"
EXTRA_OECONF += "--with-glib --with-common-includes=${STAGING_INCDIR}"
do_configure_append() {
echo "/*This is compiled to generate, only look don't try*/" > ${S}atfwd_config.h
echo "#ifndef _ATFWD_CONFIG_H_" >> ${S}atfwd_config.h
echo "#define _ATFWD_CONFIG_H_" >> ${S}atfwd_config.h
#//<!-- ODM feature caogang#2015-07-13
if [ "${PRJ_NAU8810}" = "NAU8810_CODEC" ]; then
echo "#define NAU8810_CODEC" >> ${S}atfwd_config.h
fi
if [ "${FEATURE_ACDB_ENABLE}" = "true" ]; then
echo "#define FEATURE_ACDB_ENABLE 1" >> ${S}atfwd_config.h
fi
if [ "${PRJ_XXX}" != "" ]; then
echo "#define ${PRJ_XXX}" >> ${S}atfwd_config.h
fi
#//end-->
The onenet.bb is:
inherit pkgconfig cmake
DESCRIPTION = "onenet sdk"
LICENSE = "PD"
PR = "r0"
LIC_FILES_CHKSUM = "file://${WORKDIR}/git/LICENSE;md5=bae84cdd023be37582157d865da54cc6"
SRCREV = "065d98dd8de91544315d6167ce73626ce739666d"
SRC_URI = "git://github.com/cm-heclouds/MQTT.git;protocol=https"
S = "${WORKDIR}/git/mqtt_sdk"
do_install() {
install -d ${D}/usr/lib
install -d ${D}/usr/include/onenet
install -m 0644 ${B}/bin/libmqtt.so -D ${D}/usr/lib/
for inc in $(find ${S} -name *.h ! -name 'cJSON.h'); do
install -m 0644 ${inc} -D ${D}/usr/include/onenet
done
}s
The sanity check documentation explains this:
dev-deps: Checks that all packages except -dev or -staticdev packages
do not depend on -dev packages, which would be a packaging bug.
It's telling you that in your current recipe "XXX" runtime-depends on "AAA-dev" and that this is a normally an error. You need to find out how/why this dependency is added before you can decide what the correct solution is.
Based on the added recipes: The issue seems to be that onenet build produces an unversioned ".so" file. This is typically a mistake (the actual library file should be e.g. "libmqtt.so.1.1" and the unversioned file should just be a symlink to the versioned one). I'm very surprised that you are not getting a fatal error on this issue when you build onenet. Are you suppressing the QA error for this?
Since you've managed to build onenet somehow, you now probably have a onenet-dev package that erroneously contains the actual library: The build system notices this during atfwd-daemon build, adds a runtime dependency to onenet-dev (because that's where the library is) and then the QA error triggers because normal packages should not depend on -dev packages.
Possible fixes:
Either fix the onenet build system so it produces a versioned library, or
Force the .so file to be packaged into the actual onenet package instead of onenet-dev, like this:
FILES_${PN}-dev = "${includedir}/"
FILES_${PN} += "${libdir}/libmqtt.so"
A bonus suggestion: Using directory variables instead of paths like /usr/include and /usr/lib (like I did above) is a good "Best Practice".
I'm trying to modify an existing zsh prompt to work with zsh 5.0 and 4.3 because those the versions the systems that I use. How would I make a zsh-script be aware of the current working directory instead of the directory that the file is in?
For context,
This is a function in the script that checks if we're currently in a git directory and adds to the prompt if we are:
# Git status.
# Collect indicators, git branch and pring string.
spaceship_git_status() {
[[ $SPACESHIP_GIT_SHOW == false ]] && return
# Check if the current directory is in a Git repository.
command git rev-parse --is-inside-work-tree &>/dev/null || return
# Check if the current directory is in .git before running git checks.
if [[ "$(git rev-parse --is-inside-git-dir 2> /dev/null)" == 'false' ]]; then
# Ensure the index is up to date.
git update-index --really-refresh -q &>/dev/null
# String of indicators
local indicators=''
indicators+="$(spaceship_git_uncomitted)"
indicators+="$(spaceship_git_unstaged)"
indicators+="$(spaceship_git_untracked)"
indicators+="$(spaceship_git_stashed)"
indicators+="$(spaceship_git_unpushed_unpulled)"
[ -n "${indicators}" ] && indicators=" [${indicators}]";
echo -n " %Bon%b "
echo -n "%{$fg_bold[magenta]%}"
echo -n "$(git_current_branch)"
echo -n "%{$reset_color%}"
echo -n "%{$fg_bold[red]%}"
echo -n "$indicators"
echo -n "%{$reset_color%}"
fi
}
However, based on my debugging, it appears that the function always believes that it is in the directory from which the script was sourced. In other words, as I change directory, the script continues to reference the directory where the script is located.
The spaceship_git_status function is called here:
# Build prompt line
spaceship_build_prompt() {
spaceship_host
spaceship_current_dir
spaceship_git_status
spaceship_nvm_status
spaceship_ruby_version
spaceship_venv_status
}
And this is the PROMPT environment variable is:
# Compose PROMPT
PROMPT=''
[[ $SPACESHIP_PROMPT_ADD_NEWLINE == true ]] && PROMPT="$PROMPT$NEWLINE"
PROMPT="$PROMPT $(spaceship_build_prompt) "
[[ $SPACESHIP_PROMPT_SEPARATE_LINE == true ]] && PROMPT="$PROMPT$NEWLINE"
PROMPT="$PROMPT $(spaceship_return_status) "
I think this is an issue with zsh versions < 5.2 because the prompt renders fine on my other computer with 5.2.
Full code: https://github.com/denysdovhan/spaceship-zsh-theme/blob/master/spaceship.zsh
I'm using Travis for continuous build and integration.
after_success:
- sbt publish
While we want Travis to build all of our branches and pull requests, we only want it to publish when on master branch.
publishTo <<= version { (v: String) =>
val nexus = s"asdf"
/* Don't let Travis publish when building pull requests.
* $TRAVIS_PULL_REQUEST == "false" if it's not a pull request. So we wan't publishTo to be
* None when TRAVIS_PULL_REQUEST != false.
*/
if(Try(sys.env("TRAVIS_PULL_REQUEST")).getOrElse("false") != "false")
None
/* Don't let Travis publish except when building master. */
if(Try(sys.env("TRAVIS_BRANCH")).map(_ != "master").getOrElse(false))
None
else if(v.trim.endsWith("SNAPSHOT"))
Some("snapshots" at nexus + "snapshots")
// don't let Travis publish releases, either
else if(Try(sys.env("TRAVIS")).getOrElse("false") == "true")
None
else
Some("releases" at nexus + "releases")
})
The problem with this approach is that Travis compiles branches twice because it fails publishing at the very last step.
How can Travis be completely prevented from running sbt publish when on non-master branch?
You might consider handling this outside of your publish script:
after_success:
- test $TRAVIS_PULL_REQUEST == "false" && test $TRAVIS_BRANCH == "master" && sbt publish
I'd make a new task for it, and call that in Travis.
val publishMasterOnTravis = taskKey[Unit]("publish master on travis")
def publishMasterOnTravisImpl = Def.taskDyn {
import scala.util.Try
val travis = Try(sys.env("TRAVIS")).getOrElse("false") == "true"
val pr = Try(sys.env("TRAVIS_PULL_REQUEST")).getOrElse("false") == "true"
val branch = Try(sys.env("TRAVIS_BRANCH")).getOrElse("??")
val snapshot = version.value.trim.endsWith("SNAPSHOT")
(travis, pr, branch, snapshot) match {
case (true, false, "master", true) => publish
case _ => Def.task ()
}
}
publishMasterOnTravis := publishMasterOnTravisImpl.value
publishTo := {
val nexus = s"asdf"
if (version.value.trim.endsWith("SNAPSHOT"))
Some("snapshots" at nexus + "snapshots")
else
Some("releases" at nexus + "releases")
}
The key here is using Def.taskDyn, which lets me call publish at the tail position without depending on it. publishTo would be plain logic.
I want to be able to tell if a command exists on any POSIX system from a shell script.
On Linux, I can do the following:
if which <command>; then
...snip...
fi
However, Solaris and MacOS which do not give an exit failure code when the command does not exist, they just print an error message to STDOUT.
Also, I recently discovered that the which command itself is not POSIX (see http://pubs.opengroup.org/onlinepubs/9699919799/idx/utilities.html)
Any ideas?
command -v is a POSIX specified command that does what which does.
It is defined to to return >0 when the command is not found or an error occurs.
You could read the stdout/stderr of "which" into a variable or an array (using backticks) rather than checking for an exit code.
If the system does not have a "which" or "where" command, you could also grab the contents of the $PATH variable, then loop over all the directories and search for the given executable. That's essentially what which does (although it might use some caching/optimization of $PATH results).
One which utility is available as shell script in the Git repository of debianutils package of Debian Linux. The script seems to be POSIX compatible and you could use it, if you take into account copyright and license. Note that there have been some controversy whether or not and how the which utility should be deprecated; (at time of writing) current version in Git shows deprecation message whereas an earlier version added later removed -s option to enable silent operation.
command -v as such is problematic as it may output a shell function name, an alias definition, a keyword, a builtin or a non-executable file path. On the other hand some path(s) output by which would not be executed by shell if you run the respective argument as such or as an argument for command. As an alternative for using the which script, a POSIX shell function using command -v could be something like
#!/bin/sh
# Argument $1 should be the basename of the command to be searched for.
# Outputs the absolute path of the command with that name found first in
# a directory listed in PATH environment variable, if the name is not
# shadowed by a special built-in utility, a regular built-in utility not
# associated with a PATH search, or a shell reserved word; otherwise
# outputs nothing and returns 1. If this function prints something for
# an argument, it is the path of the same executable as what 'command'
# would execute for the same argument.
executable() {
if cmd=$(unset -f -- "$1"; command -v -- "$1") \
&& [ -z "${cmd##/*}" ] && [ -x "$cmd" ]; then
printf '%s\n' "$cmd"
else
return 1
fi
}
Disclaimer: Note that the script using command -v above does not find an executable whose name equals a name of a special built-in utility, a regular built-in utility not associated with a PATH search, or a shell reserved word. It might not find an executable either in case if there is non-executable file and executable file available in PATH search.
A function_command_exists for checking if a command exists:
#!/bin/sh
set -eu
function_command_exists() {
local command="$1"
local IFS=":" # paths are delimited with a colon in $PATH
# iterate over dir paths having executables
for search_dir in $PATH
do
# seek only in dir (excluding subdirs) for a file with an exact (case sensitive) name
found_path="$(find "$search_dir" -maxdepth 1 -name "$command" -type f 2>/dev/null)"
# (positive) if a path to a command was found and it was executable
test -n "$found_path" && \
test -x "$found_path" && \
return 0
done
# (negative) if a path to an executable of a command was not found
return 1
}
# example usage
echo "example 1";
command="ls"
if function_command_exists "$command"; then
echo "Command: "\'$command\'" exists"
else
echo "Command: "\'$command\'" does not exist"
fi
command="notpresent"
if function_command_exists "$command"; then
echo "Command: "\'$command\'" exists"
else
echo "Command: "\'$command\'" does not exist"
fi
echo "example 2";
command="ls"
function_command_exists "$command" && echo "Command: "\'$command\'" exists"
command="notpresent"
function_command_exists "$command" && echo "Command: "\'$command\'" does not exist"
echo "End of the script"
output:
example 1
Command: 'ls' exists
Command: 'notpresent' does not exist
example 2
Command: 'ls' exists
End of the script
Note that even the set -eu that turns -e option for the script was used the script was executed to the last line "End of the script"
There is no Command: 'notpresent' does not exist in the example 2 because of the && operator so the execution of echo "Command: "\'$command\'" does not exist" is skipped but the execution of the script continues till the end.
Note that the function_command_exists does not check if you have a right to execute the command. This needs to be done separately.
Solution with command -v <command-to-check>
#!/bin/sh
set -eu;
# check if a command exists (Yes)
command -v echo > /dev/null && status="$?" || status="$?"
if [ "${status}" = 127 ]; then
echo "<handle not found 1>"
fi
# check if a command exists (No)
command -v command-that-does-not-exists > /dev/null && status="$?" || status="$?"
if [ "${status}" = 127 ]; then
echo "<handle not found 2>"
fi
produces:
<handle not found 2>
because echo was found at the first example.
Solution with running a command and handling errors including command not found.
#!/bin/sh
set -eu;
# check if a command exists (No)
command -v command-that-does-not-exist > /dev/null && status="$?" || status="$?"
if [ "${status}" = 127 ]; then
echo "<handle not found 2>"
fi
# run command and handle errors (no problem expected, echo exist)
echo "three" && status="$?" || status="$?"
if [ "${status}" = 127 ]; then
echo "<handle not found 3>"
elif [ "${status}" -ne 0 ]; then
echo "<handle other error 3>"
fi
# run command and handle errors (<handle not found 4> expected)
command-that-does-not-exist && status="$?" || status="$?"
if [ "${status}" = 127 ]; then
echo "<handle not found 4>"
elif [ "${status}" -ne 0 ]; then
echo "<handle other error 4>"
fi
# run command and handle errors (command exists but <handle other error 5> expected)
ls non-existing-path && status="$?" || status="$?"
if [ "${status}" = 127 ]; then
echo "<handle not found 5>"
elif [ "${status}" -ne 0 ]; then
echo "<handle other error 5>"
fi
produces:
<handle not found 2>
three
./function_command_exists.sh: 34: ./function_command_exists.sh: command-that-does-not-exist: not found
<handle not found 4>
ls: cannot access 'non-existing-path': No such file or directory
<handle other error 5>
The following works in both bash and zsh and avoids both functions and aliases.
It returns 1 if the binary is not found.
bin_path () {
if [[ -n ${ZSH_VERSION:-} ]]; then
builtin whence -cp "$1" 2> /dev/null
else
builtin type -P "$1"
fi
}