Use relative paths to link library file - xcode

My xcode project links to the TestFlight library (libTestFlight.a) and the project is also under git (locally) which I then push up to bitbucket.
I then pull the project down to my macbook pro. However the first build fails because its trying to find the libTestFlight.a file using a path that is relative to my iMac. I can change the project settings and all is OK, but I'd like to avoid this happening on every refresh between machines.
I can't seem to find a way to make the linker find the file relative to the xcode project directory.

If your bundle contains the static library, then in Build Settings you can put the value
"$SOURCE_ROOT"
No matter which machine your code runs in, it will always point to the correct root folder.
The same can be given for Header Search Paths.

Related

Xcode 13 and Linux Makefile

I searched and it seems all the answers are from are from many years ago. I'm working on a command line tool for linux/Mac that currently uses a Makefile. I wanted to use the Xcode debugger, so I tried to build an Xcode project, but I'm having some final difficulties.
I created a new project of type command line tool with external build target in the project directory. The default was to use /usr/bin/make. It ended up in a subdirectory, and based on several older web resources, I moved the project file up to the folder with the makefile. I added the source files in the folder to the project and right now I can edit and build from Xcode without difficulty.
The issue is that the final executable ends up in a subfolder called bin, and I can't seem to discover how to tell Xcode that that is the final executable. All of the resources I've found talk about adding a custom executable to the target from the project menu, but that menu no longer exists. Searching project help doesn't seem to point to any setting for the executable produced by an external build process. The closest build setting I could find was PROJECT_NAME, but changing that didn't seem to help.
Thanks.
Found an answer on another site. Edit the info tab in the scheme to identify the binary.

why is xcode 5 not finding a header file for newrelic enough though the library is there?

I added the newrelic objective-c library to my project and it worked fine. Compiled no issues. Checked everything into git and did a git pull from another machine. It compiles fine on that machine too.
But on our QA machine it gives the error
'NewRelicAgent/NewRelicAgent.h' file not found.
The QA machine is also using xcode 5 but there must be some difference I can't figure out.
The library is there under Frameworks. The header file is there I can see it! Any ideas?
One small difference I noticed is on my machine in the finder the icon for the library is:
A lego peice. But on the QA machine it's just a normal folder icon.
How did you add the framework? If you added it to the Build Phases pane, the headers dir should have been added to the header search path.
I have found that for whatever reason, when I add a third party framework, it works if you make a folder, not just a group, and then put the framework file in there. So the flow would be:
Make a group called thirdparty
Go to the inspector and make sure that the checkbox is checked for it to use a folder, and that there is a real folder there
Add it on the build phases pane, pointing to it in the folder
Then try and use an import. If it doesn't work, double check that the header dir was added. I did this many times with OCHamcrest.

Qt5 build has absolute paths

I'm putting together a Qt5.1 build for our team to use for our project. I was able to build 32 and 64 bit binaries with debug symbols using the instructions here.
I built using VS2012.
After finishing the build I copied the bin, lib and include folders to our resources folder and all was well. However, a couple weeks later I moved the folder I had built from, and my application stopped running giving me the error.
I thought this was odd so I began digging around. I found that there were many references in my build to the original build directory.
I could just keep the build folder the same as it was before or point the compiler to the original directories, however, on other people's machines this might not work if they don't have the same drive letters.
Is there any way to build Qt5 so that it does not use absolute paths? I imagine there must otherwise nobody would be able to distribute it.
You need to add a qt.conf file in your BIN directory.
[Paths]
Prefix = $(QTDIR)
Then set an environment variable called QTDIR with the path to your QT files.

Xcode 4 and nested projects -- header files not found

I'm having a myriad of problems with Xcode 4 and nested projects that worked just well under Xcode 3.2. Here's a very basic one I cannot solve:
I'm building a cocoa framework that requires another cocoa framework for which I have the source. So I perform the usual steps:
Drag the .xcodeproj file of the required framework into my main framework project
In my main framework under TARGETS > MyFramework > Build Phases > Target Dependencies: Add the nested project's target
Make sure the header files of the nested framework are public
In Xcode Settings > Locations > Build Location I have it set to Place build products in derived data location (recommended)
Build products path of both targets are set to ${BUILT_PRODUCTS_DIR} and tell me they are at the DerivedData/Debug (or Release) location
Architecture settings for both targets are identical
Then I hit [CMD] + B to build and it tells me that it doesn't find the header files of the nested framework. When I check the settings, User Header Search Paths contain the path to DerivedData/Debug, and inside there is the nested framework target with the header files in Versions/A/Headers.
I'm sitting here, anybody an idea what I'm doing wrong?
The issue goes away when building for Debug when I change the User header search paths to ${BUILT_PRODUCTS_DIR}/MyFramework.framework/Headers. However this doesn't work when building for Distribution as the frameworks then use their Release settings, which ends up in a different subdirectory...
My temporary solution is to also define a Distribution configuration for the nested projects. This way the headers are found and the linker can link successfully.
Here's my synthesized knowledge so far:
Forget the whole public header thing with Xcode, it's a PITA and doesn't work correctly when archiving your app. Instead, have all static library header files on the project level and tell your app where to find it.
Ease your pain by making sure all targets have the same name for the build configuration (i.e. add an "AdHoc" and "Deployment" configuration to the static libraries).
In build settings, point the Header Search Paths (if you use #include <file.h>) or User Header Search Paths (if you use #include "file.h") to the directory of the static library project. If the static library project is inside your app directory, use this:
"$(PROJECT_DIR)" (recursive enabled)
If you have a directory which contains a) the static library project and b) your app, then this should work:
"$(PROJECT_DIR)/.." (recursive enabled)
If the submodule contains compiled libraries, set your Library Search Paths to:
"$(TARGET_BUILD_DIR)"
Make sure all the static library projects that you use have Skip Install set to YES.
Again, no public header files (Build Phases » Copy Headers) in any of the static libraries, otherwise Xcode will not be able to archive an app.
Make sure to tell Xcode when to build the static libraries, as shown in this Tech Doc from Apple.
Old Answer:
I still haven't found a real solution to this problem with static libraries. What works for me is:
Create an "AdHoc" Configuration for the static library
Add $(BUILT_PRODUCTS_DIR) to User Header Search paths for the application (with recursive checked) -> this is used when running the app
In the Xcode menu, select Product > Build For > Build For Archiving
This works, the app finds the header files and builds itself, it ends up in DerivedData//Build/Products/AdHoc-iphoneos/ as an App bundle. Following these simple instructions (dead link) from TestFlightApp.com I can pack this App into an IPA and send it around. Simply selecting to Archive the app from Xcode does again not find the headers, even if they truly are in the AdHoc-iphoneos build directory.
(As of Xcode 5.1)
When the subproject is built by XCode, the subproject header files are copied into the build directory. When archiving, it seems that this copy destination directory is not added to the header/include search path. You'll want to go to your Build Settings and add
$(BUILD_ROOT)/../IntermediateBuildFilesPath/UninstalledProducts/include
to the "Header Search Paths" for the scheme that you use for archiving.
If you're not sure which scheme is used for archiving, go to Product -> Scheme -> Edit Schemes and look for Archive in the left column.
Make sure your third party framework is added as «group» to your main project, so you can see it in your project's hierarchy...
I had the same problem here and I could solve the problem by setting "Build Location" to Place build products in locations specified by targets"
I had this problem: I could build both Debug and App Store configurations, but not Ad Hoc. Building Ad Hoc gave me errors because it couldn't find .h files needed by nested projects.
Turned out I had an expired provisioning lingering in my Release configuration. I updated that provisioning link and now I can both build Ad Hoc and use the Archive feature to package it.
Took me hours to figure it out! My mind just didn't jump from missing .h files to provisioning errors just by itself. =) There might have been an error or warning complaining about the missing provisioning, but if so it was well buried among the hundreds of .h related errors.
I was having the same issue with a Configuration named "Ad Hoc" (as per TestFlight recommendation at http://help.testflightapp.com/customer/portal/articles/402782-how-to-create-an-ipa-xcode-4) and the main project could not find some of the headers from the nested projects. I renamed the project to "AdHoc" (no spaces) and the problem went away; seems like spaces can mess up header search paths in some cases, although I haven't figured out the specifics of when that might happen and why.
I was having this issue with a nested project that built a static library. I found this doc on apples site that completely saved my life.
http://developer.apple.com/library/ios/#DOCUMENTATION/Xcode/Conceptual/ios_development_workflow/AA-Developing_a_Static_Library_and_Incorporating_It_in_Your_Application/archiving_an_application_that_uses_a_static_library.html
I'm so glad I didn't have to muck around with the derived data paths.
For me, this happened after a GIT merge, which created many conflicts, one of them related to the project file. After the merge, I'm sure the structure of the project file changed.
What I ended up doing was going into the project "Build Settings", then looking for "Always Search User Paths" and turning it to Yes.
I guess the merge turned this boolean to No, therefore the project wasn't looking in the right places for the header files.

How to transfer Eclipse workspace and project from Windows to Linux and Mac

We have a a product developed on Windows for years. The product is composed of one Eclipse workspace and about 20 projects.
On Windows, we ask every developer check out projects into d:\dev\product folder, and copy a unified Workspace to d:\dev\prod_workspace. This way, whenever a new machine is set, we simply copy files to the same folder, and we can start working immediately.
Now we need to move our development environment to Linux and Mac. But there's no D:\ on Unix. And home folder for Linux is mostly like /home/username and /Users/username for Mac. We found Eclipse keeps absolute path in workspace when referring to projects, so simply copy workspace over does not work anymore. Even when we manually create/configure workspace on a Linux machine, it still cannot be copied over to another user, because the absolute path is changed.
I guess our goal is to allow easy setup of development environment. Do you have any suggestion to move eclipse workspace around?
I develop an Eclipse based product for Mac and WIndows (haven't tried Linux).
The solution I found to work best is to actually go and manually check out the projects in the workspace on each machine directly from source control. While the project structure does convert between platforms, any attaches, such as version-control stuff does not. I am not sure why, but I guess each thing has its kinks.
You may be able to able to hasten things a little by creating a project working set (or whatever it's called, I think it's a PSF file) for a platform, and then reuse it on all platforms.
Another problem is that Eclipse versions are not 100% compatible. One of the problems I have is with manifest files for plugins, which have different semantics (e.g., what to do with exported packages that don't actually exist) in each platform, causing a headache.
Finally, be away that Java on Mac and PC are not identical. In fact, Eclipse has two versions for the Mac. I usually end up running and compiling on Java 5 on the mac, which does have some incompatibilities with Java 6.
I have ported my eclipse project from windows xp to RHEL(never tried Mac)
Your task can be accomplished in the following steps :-
Use workspace variables in your project code rather than absolute path.
Shift your workspace to some location in linux as /Workspace or if you want to keep it user specific make the workspace folder as /home/user_name/Workspace.
There is an option to change all the \ of windows to / in linux. The option can be found under the file menu in eclipse.
Change the settings of where to find classpath directories, the options can be found by just right clicking on the project menu in the project viewer panel.
There is an option to build the project clean from scratch. You just have to click on the option and eclipse would re-compile the project.
Once all this is done, and all your database connections have been successfully ported to linux, you would have a working project running in linux as well.
I can think of two ways to do this:
Use workspace variables in Eclipse to point to the exact location where the workspace sits, if you need that. Then the developer can put everything in the workspace on Windows, Unix, whatever, then define the variable and you're done, or
If compilation is always done from within Eclipse, meaning you don't truly need any absolute paths anywhere, then change the projects to all use relative paths and check the Eclipse classpaths, etc, into source control. Thus, when a developer checks out everything in Eclipse, the classpath and .project file will be at the root level in the project and all paths will be pre-defined.
For option #2, you may need to have multiple .project or .classpath files, and have the developer copy the appropriate one into location. That is, copy .classpath.win into .classpath for Eclipse's use.
If you check out the projects from a source repository, then consider using Team -> ProjectSets to handle these.
Preferences can be saved in a file and loaded.
We tried classpath variables and loathed them. Now we just have everything in single projects in the source repository.
Use the Workspace Preference Transferrer plugin
This allows you to transfer the workspace settings from a given workspace to another via new options in the 'switch workspaces' dialog.
I'll share my own solution:
Zip the workspace. Just zip the entire file and then transfer it over. Then, all you need to do is delete the default drive reference:D:\ ...or whatever ends up appearing in the beginning of your file name. Then all you need to do is drag it over to the new empty workspace on your newly downloaded Eclipse.

Resources