How to use modules replace functionality in cloud functions - go

I have a google cloud function that is a subdirectory in a repository. It uses the "Directory with source code" option in the settings menu. I keep getting this error on deploy:
Deployment failure:
Build failed: go: parsing /utils/pubsub/go.mod: open /utils/pubsub/go.mod: no such file or directory
go: error loading module requirements
I'm assuming that GCF does not upload the entire directory to the instance, but instead only the folder? This breaks the replace functionality of Go modules. Is there something I am doing wrong?
Link to the repo: https://github.com/FreekingDean/jeffbotgo/tree/5d735cc/slackevent

I work at Google and on this product.
Only the directory where you run gcloud is uploaded. There is no staging step beyond zipping the current directory and uploading it.
Notably, modules are preferred by the builder over vendor. If there is a go.mod, modules will be used. When you upload your function, it only includes the directory with your function at the root, not any directories one level up. So, when there is a go.mod and you have a replace directive pointing one level up, it will not work.
The solution for now with this layout is to vendor and not upload the go.mod/go.sum files. When using gcloud, you can create a .gcloudignore file to do this for you. See https://cloud.google.com/functions/docs/concepts/go-runtime#specifying_dependencies for more detail. Alternatively, modify your project to include any necessary helper packages in subdirectories.

I had the same issue today.
When reading thru the documentation for the 8th time i came across a warning box bellow the "Vendor directory" headline.
Warning: If your project has both a go.mod file and a vendor directory
at the root of your project, the vendor directory will be ignored
during deployment. You must use a .gcloudignore file to ignore the
go.mod file in order to ensure that your vendor directory is used
during deployment.
So basically once i added a .gcloudignore file with go.mod (will add go.sum as well) everything worked. So i guess if you have a go.mod file the cloud function will try to fetch dependencies instead of using the ones uploaded in the vendor folder.
I'm just guessing here tough.

Related

VS Code showing me "Error loading workspace: found module "main.go" twice in the workspace"

I am using the primary GO extension.
I use VS code a lot, now I'm learning GO lang.
And when I open VS Code every time I'm getting this:
Error loading workspace: found module "main.go" twice in the workspace.
While running the code it's giving the right output.
I don't know how to fix this.
Anybody, help me with this error.
It would be better to open in VSCode only one folder with its own go.mod project.
A workspace with multiple go.mod/project should be supported with 1.18
The go command now supports a "Workspace" mode.
If a go.work file is found in the working directory or a parent directory, or one is specified using the -workfile flag, it will put the go command into workspace mode.
In workspace mode, the go.work file will be used to determine the set of main modules used as the roots for module resolution, instead of using the normally-found go.mod file to specify the single main module.
As described in "How to make VScode Go work in a Multi-Module Repo" from Varun Kumar, this used to work:
If you want to work with all the nested modules in a single workspace, there is an opt-in module feature that allows to work with multiple modules without creating workspace folders for each module. Set this in your settings -
"build.experimentalWorkspaceModule": true
But as per september 2022 is deprecated.
See more at gopls documentation "Setting up your workspace".

How to use an alternate go.mod file for local development?

Currently I am working on an API which uses Serverless Framework with Go.
I'm using the Serverless-offline plugin for local testing.
This API depends on a few other repositories (which I also maintain), which I import using the go.mod file.
However I am having a hard time refining my developer workflow.
Currently, if I want to make changes in a repository which this API depends upon, I have to alter the projects go.mod to include replace directives for the purpose of testing, but then I'm having to manually change it back for deployment to production.
Basically I'm looking for a way to include replace directives, which only get applied during local development. How has everyone else dealt with this problem?
Bonus question: Is there any way to run Serverless offline in docker? I'm finding that serverless-offline running on the bare metal is causing inconsistencies between different developers environments.
You can run go commands with an alternate go.mod file with the -modfile option:
From Build commands:
The -modfile=file.mod flag instructs the go command to read (and
possibly write) an alternate file instead of go.mod in the module root
directory. The file’s name must end with .mod. A file named go.mod
must still be present in order to determine the module root directory,
but it is not accessed. When -modfile is specified, an alternate
go.sum file is also used: its path is derived from the -modfile flag
by trimming the .mod extension and appending .sum.
Create a local.go.mod file with the necessary replace directive for development and build, for example, with:
go build -modfile=local.go.mod ./...

How can I use a local repository in the vendor in Google Cloud Functions with Go

I'm trying to deploy a Google Cloud Function written in Go.
By doing some research I found out that vendor files are prefered over go.mod so I'm vendoring everything I use (which includes some local dependencies) and ignoring the go.mod/sum files in the .gcloudignore file.
The problem is that after trying to deploy, I get the following error:
go: nimbus#v0.0.0-00010101000000-000000000000: parsing /nimbus/go.mod: open /nimbus/go.mod: no such file or directory; Error ID: 03a1e2f7
nimbus is my local dependency and it has the following structure:
My Function repository has the following structrure:
and my go.mod file is:
module my_function
go 1.13
require nimbus v0.0.0-00010101000000-000000000000
replace nimbus => ../../../nimbus
I've tried this solution https://stackoverflow.com/questions/5441096 already. But it did not fix my issue.
I've tried everything to solve this issue, but nothing seems to work.
If you have a go.mod file and a vendor directory, the vendor directory will be ignored when you deploy your function.
https://cloud.google.com/functions/docs/writing/specifying-dependencies-go
I have used modules when deploying GCP functions in Go. Haven't had any problems. But I can't speak to the preference of using vendor/ instead. It should work, just without the go.mod file.
Turns out the problem was very complicated and I hope Google finds a solution for it asp.
By deploying my function using Cloud Build, It would read from my repository on Google Source, but, by reading from there it would bypass the .gcloudignore file and deploy both the go.mod/sum files and the vendor directory with my local code.
As said in https://stackoverflow.com/a/62050872/10316247:
If you have a go.mod file and a vendor directory, the vendor directory will be ignored when you deploy your function.
So the error would occur because of my go.mod not being able to find local repository.
My solution was to rename my go.mod/sum files so it would not be considered:
When you use golang 1.16 and specify golang 1.16 in the go.mod folder it will instead default to using the vendor files with the --mod=vendor flag set, which will solve this issue.
You'll just need to ensure your module name is formatted correctly (something like example.com/module).

Why is go-get trying to download local code from a remote location?

I recently added a new package and directory to my Go project, and now when I try to build it, I get errors about a password error on Gitlab.
I am not importing a new remote package, I am simply adding a new directory underneath my already declared module path. For instance, my go.mod has gitlab.com/example/api and the package I added is gitlab.com/example/api/postgres.
I am not actually hosting on gitlab, I just needed something to name the project as I worked on it. Clearly it won't find it on gitlab, but it is available locally. Why is go-get trying to download a package/path that is available locally?
Why is it only happening for this new package, and not for all of the existing package under this path?
Golang 1.14
You have to add replace above the require block in your go.mod to work with local package. For example:
replace gitlab.com/example => /Users/abc/projects/gitlab.com/example
Ref: https://github.com/golang/go/wiki/Modules

Automatic Extension Update: Unknown Archive type

I have created a new version of my Joomla extension.
Manual upgrade via zip file or the directory works fine.
But the automatic upgrade (which used to work fine before), now gives an error 500 and the following error messages: "Unknown Archive type", "*Update path does not exist" and "Installation unexpectedly terminated: Update path does not exist".
I have no idea why those messages appear.
The update.xml references the correct zip files. Downloading it manually works just fine.
Joomla(/php/apache) has all rights on the folder containing the joomla installation.
After trying the automatic update, the tmp folder contains the downloaded zip archive with the latest extension version, interestingly without the .zip extension. Is the Joomla downloader not correctly naming the file and then failing upon finding that the file doesn't have a .zip extension?`Or what could it be?
Would be very thankful for any ideas...
Edit: My project is hosted on github, and github seems to automatically create a subfolder in the downloaded zip archive, named -.zip.
I'm using a link to the tagged github zip directly in my update.xml
I'm not sure if github always added this folder in the zip file, back when it still worked for me...
Might the Joomla problem have to do with the zip file containng such a folder, and not directly the extension stuff at root level? If so, anybody know if/how I can change github to not create that subfolder?
Right, just had a quick test of this.
I couldn't seem to find out how to automatically zip up a sub folder (there is a way but I need to do some more research/ask questions regarding this), however what you can do is the following:
Create a zip of your Repo
Open the zip, extract the folder you wish to be zipped then zip it
Create a new version and then drag your zip file into the upload box
Publish the release
Here is an example, have a look at the "Creating Releases" sections at the bottom:
https://github.com/blog/1547-release-your-software
Hope this helps
To answer my own question:
Yes, github seems to have recently changed their policy to create a root folder in the zip file, named as the repository the zip file is downloaded for (stupid, if you ask me, since the exact same information is encoded in the zip file name already anyway!).
Edit and Rewrite: It seems that either something changed in Joomla or that if you adhere to a naming convention - namely the root folder in the zip file having the exact extension name (or, I think and have to test, actually the same as the file name, without the version information), then the automatic update will work.
So as in my case, I have a Joomla package; the package is now in a repository pkg_mypkg. The zip file generated by github has the name pkg_mypkg-version.zip (e.g. pkg_myfancyext-1.0.9.zip), and contains a folder named pkg_mypkg. And inside the pkg_mypkg folder is a pkg_mypkg.xml file, the extension manifest. And this actually seems to be the configuration where automatic update works.

Resources