Update CI/CD notes.

Signed-off-by: Munyoki Kilyungi <me@bonfacemunyoki.com>

1 files changed, 16 insertions, 24 deletions

diff --git a/issues/CI-CD/failing-services-startup.gmi b/issues/CI-CD/failing-services-startup.gmi
index 169bccf..f820b72 100644
--- a/issues/CI-CD/failing-services-startup.gmi
+++ b/issues/CI-CD/failing-services-startup.gmi
@@ -11,7 +11,7 @@
 
 ## Description
 
-On rebuild of the CI/CD container with guix channel pinned at commit `34453b97005ff86355399df89c8827c57839d9c7`, some services fail to start and the error messages we get are as follows:
+Upgrading guix to `34453b97005ff86355399df89c8827c57839d9c7` for CI/CD fails with:
 
 ```
 2025-08-20 16:05:20 Backtrace:
@@ -41,43 +41,35 @@ On rebuild of the CI/CD container with guix channel pinned at commit `34453b9700
 2025-08-20 16:05:20 Wrong type (expecting exact integer): #f
 ```
 
-The services that fail are:
+Failing services:
 
 * genenetwork3: consistently
 * genenetwork2: consistently
 * gn-auth: intermittently
 
-After digging further into this issue, and I think I have the beginnings of an idea of why the issue is comming up. Looking at:
 
-=> https://codeberg.org/guix/guix/src/commit/34453b97005ff86355399df89c8827c57839d9c7/guix/build/syscalls.scm#L1218-L1233
+## Troubleshooting Notes
 
-We see the documentation says:
-
-> Note that CLONE_NEWUSER requires that the calling process be single-threaded,
-> which is possible if and only if libgc is running a single marker thread; this
-> can be achieved by setting the GC_MARKERS environment variable to 1.  If the
-> calling process is multi-threaded, this throws to 'system-error' with EINVAL.
-
-and looking at the error we are getting:
+To reproduce, set the default directory for all GN cloning to 'gn-repositories':
 
 ```
-⋮
-2025-08-20 15:17:38 guix/build/syscalls.scm:1231:10: In procedure unshare: 268566528: Invalid argument
-⋮
+@@ -775,38 +775,43 @@ described by CONFIG, a <genenetwork-configuration> object."
+                 #~(make-forkexec-constructor
+                    (list #$(least-authority-wrapper
+                             (program-file "genenetwork2"
+                                           (genenetwork2-cd-gexp config))
+                             #:name "genenetwork2-pola-wrapper"
++			    #:directory gn-repositories
 ```
 
-Now, looking at
-=>https://codeberg.org/guix/guix/src/commit/34453b97005ff86355399df89c8827c57839d9c7/gnu/build/linux-container.scm#L321 where `unshare` is called
-
-
-we could come to the conclusion that, perhaps the calling process for `unshare` for "genenetwork3" and "genenetwork2" is consistently multi-threaded, leading to the error above.
-
-It might also explain why the /gn-auth/ service will **sometimes** throw the same error when the container is restarted, but other times, it'll just start with no error.
+On the first run, gn3/gn-auth/gn-guile start normally.  gn2 fails (still under investigation).
 
-I (currently) have no idea why the calling process would be multi-threaded.
+After restarting the container, gn3/gn-auth/gn-guile fail with above error.
 
-Or maybe, I'm overthinking this whole thing.
+Removing "gn-repositories" and restarting the container makes everything run normally.
 
 ## Also See
 
 => https://issues.guix.gnu.org/78356 Broken system and home containers
+=> https://codeberg.org/guix/guix/src/commit/34453b97005ff86355399df89c8827c57839d9c7/guix/build/syscalls.scm#L1218-L1233 How "unshare" is defined
+=> https://codeberg.org/guix/guix/src/commit/34453b97005ff86355399df89c8827c57839d9c7/gnu/build/linux-container.scm#L321 Where `unshare` is called