It has quit rebuilding but still pulls down each dependency. Up to 0.7.0 if I did a build and then turned off my network I could still
rebuild as long as I did not remove the dependencies. Now, even though this project is completely built all remote dependencies
are pulled. If the network is not reachable I cannot rebuild without moving all the files to project subdirectories and changing the fpm.toml file to use local packages. This occurs on any Linux system I tried. So the actual recompilation is gone but not the pull.
I can duplicate this with just

https://github.com/urbanjost/M_strings

@mercury:~/github/M_strings$ fpm build

Update: M_msg

Reinitialized existing Git repository in /home/urbanjs/venus/V600/github/M_strings/build/dependencies/M_msg/.git/
remote: Enumerating objects: 275, done.
remote: Counting objects: 100% (275/275), done.
remote: Compressing objects: 100% (140/140), done.
remote: Total 275 (delta 147), reused 221 (delta 131), pack-reused 0
Receiving objects: 100% (275/275), 1.14 MiB | 2.65 MiB/s, done.
Resolving deltas: 100% (147/147), done.
From https://github.com/urbanjost/M_msg

• branch HEAD -> FETCH_HEAD
  Project is up to date
  @mercury:~/github/M_strings$ fpm build

Update: M_msg

Reinitialized existing Git repository in /home/urbanjs/venus/V600/github/M_strings/build/dependencies/M_msg/.git/
fatal: unable to access 'https://github.com/urbanjost/M_msg.git/': Could not resolve host: github.com
cmd_build Model error: Error while fetching git repository for remote dependency

@urbanjost I cannot reproduce this issue: on my system (macOS + gfortran-12), I have:

federico@Federicos-MacBook-Pro M_strings % fpmx build
 + mkdir -p build/dependencies
Initialized empty Git repository in /Users/federico/Documents/GitHub/M_strings/build/dependencies/M_msg/.git/
[...]
demo_rpad                              done.
[100%] Project compiled successfully.
federico@Federicos-MacBook-Pro M_strings % fpmx build
Project is up to date
federico@Federicos-MacBook-Pro M_strings % fpmx build
Project is up to date
federico@Federicos-MacBook-Pro M_strings % fpmx update
federico@Federicos-MacBook-Pro M_strings % 

Is it possible that other compilers (or versions thereof) have problems with the way some of the equality comparisons are made? Strings from cache.toml are just compared as string1==string2 but maybe there are caveats with trailing spaces / new line characters? There should not be as far as the TOML is properly loaded (which it is).

My guess is you will actually need 2 flags when you're building up the dependency list.

1. Is this the cached version?
2. Is this the first time I'm encountering this dependency

That way you only update the dependency if it doesn't match the cached version and it is the first version encountered when walking the tree.

Aaah I think I found where the issue is from. Commit 1f604b4 introduced undefined behavior (removed fetch=.false. by default) in dependency%register, apparently in all the tests I've made so far that was always .true. by complete chance!

Brad @everythingfunctional what should I do with this bugfix? merge into my release branch directly, or open a new PR on fpm main?

Brad @everythingfunctional what should I do with this bugfix? merge into my release branch directly, or open a new PR on fpm main?

Proper procedure would probably be to do a separate PR, and then rebase this one once that's been merged.

Release suspended due to #875, will open a new PR for that.