#v0.17.0-llm.2
1 messages ยท Page 1 of 1 (latest)
Tag pushed. Waiting for release...
๐ค
cc @narrow juniper @left onyx @split bolt @light vine
welp
Not sure where to go from here..
looks like there's no tag at https://github.com/dagger/dagger-go-sdk
@narrow juniper did it run all this last time? It looks like it's trying to do more than a pre-release should?
I only added the one tag
on dagger/dagger
huh, looks like llm.1 also had errors
re-running
failed again
Why is that publish green, but nothing is published?
i wonder if this is related - maybe it did more than intended? was a last-minute fix after the last prerelease: https://github.com/dagger/dagger/commit/fb7a9721c691cb752a0891025a895c4b4c352f48
these 'nightly' versions look a little
- like it's not respecting the version we gave it maybe? https://v3.dagger.cloud/dagger/traces/f6c053290ed5e4a7c0e44a1fe681b071?span=b7280ac17d4ea8af&logs#b7280ac17d4ea8af:L213
don't think so - latest is still v0.16.1
yeah ok - "Default: {{ incpatch .Version }}-{{ .ShortCommit }}-nightly."
it was {{.Tag}} before but we don't want that because it has a v in front
maybe there's a way to just remove the v 
{{ trimPrefix "v" .Tag }}?
yeah looks that's a thing
will put up a pr for that, i guess it has to be merged into main and then main -> llm
@cloud nacelle https://github.com/dagger/dagger/pull/9660
Fast-track merge right?
yup
oh - DCO, will fix and then merge
@cloud nacelle merged, and merged into llm
yeah go for it ๐
I'm seeing several traces related to that commit, including several publish --dry-run...
The only one I see that looks like a "real" publish is this one: https://v3.dagger.cloud/dagger/traces/4c6a27f7d69c1d257a3ba5d39c62275d?listen=7315879e6cfeea7d
โจฏ release failed after 0s error=yaml: line 9: did not find expected key
@cloud nacelle https://github.com/dagger/dagger/pull/9661
yep - just merged into llm
@narrow juniper back to the previous error, no change apparently
hmm the versions look right now at least
wait, not really, it seems like the downstream publish jobs do expect a v prefix?
now i'm completely lost because that means the fix would be to "just" undo that commit, but surely it was made for a reason
might be best to let @left onyx pick this one up tomorrow ๐
Yup... good effort though
looking into this now
yeah a bit confused here: https://github.com/dagger/dagger/pull/9654
not sure why we did this, i intentionally set this so that we get the right format - it's releases/<version with v>/dagger_<version with v>_<platform>.tar.gz
i would suggest we revert that? the attempt with the removed nightly block is definitely wrong, see the trace here: https://v3.dagger.cloud/dagger/traces/f6c053290ed5e4a7c0e44a1fe681b071?listen=4ac4153f6d2bd1aa&listen=c45945b1e43240f7&listen=b7280ac17d4ea8af#b7280ac17d4ea8af:L214
see the -nightly suffix
waiting for @left onyx, since yea, don't really understand the logic of that pr:
Otherwise this will append v to versions when publishing to S3, which we do not want.
...i think we do want it, seeinstall.sh's output:
https://github.com/dagger/dagger/pull/9649 is also ready to review now, should mean llm.3 can get aliased, so we don't need to keep updating install instructions
This looks almost right to me. Test SDK Provision should have worked, but all other jobs should fail since we only publish the CLI & Engine, and nothing else. Skipping those jobs on a pre-release would be worth doing from my POV.
We don't want <version with v> . This comes straight from the releases S3 bucket:
yes, there are two versions
one in the directory, which shouldn't have it
but another in the file, which should
I can see that 0.17.0.-llm.2 got eventually published. I can also see the nightly version, which came after my PR, which is not right.
OK, I am going to delete this one which is incorrect: 0.17.0-0c77012cc-nightly/
I see the problem here:
oh
All of these should indeed have the v, exactly as you were saying @split bolt
i think the name_template should be .Tag
instead of .Version
So basically we don't want the v in the dir name, but we do want the v in the actual filename.
Yes.
.Version without the v in the dir name.
And .Tag in the filename.
i'll pr
I am going to delete this CLI release too since it's wrong.
Once we get all the PRs merged, let's merge main into the llm and go for another 0.17.0-llm.2 . This one should be the right one @cloud nacelle @narrow juniper
OK, all done.
that should produce the right result now ๐
Looking at this one now.
This review will take longer than expected - dropped a comment on the PR.
Do you want to merge main into llm and force push the v0.17.0-llm.2 tag @split bolt ?
sure, can do
um i've merged + pushed, but i'm actually not sure how to verify that everything's working as expected
i follow the base example in the melvin repo, but i kind of get weirdly inconsistent results? i'm not sure if that's at all expected.
holding cutting it until hopefully someone confirm it's still working properly ๐
I think that we should still cut it, and make sure that releasing works as expected. Until the Melvin repo is updated, and until https://github.com/dagger/dagger/pull/9649 is merged, the chances of someone using a "bad" pre-release are super slim.
We can always re-publish the same pre-release tag. This is a very different game from a final, stable release. Things failing is the norm in this world.
yeah okay! i'm just gonna get some lunch first.
๐ฑ
great that's not right either
the goreleaser is not joyous to edit ๐ฆ
just kinda confused why Tag doesn't exist here, but does in another context ๐ฑ
fair
welp, let's keep trying https://github.com/dagger/dagger/pull/9669
Nope, Tag is not set in this context helpfully, so instead just use v + Version here.
have deleted the tag again, will try again once above checked and released
i have manually run the goreleaser dry-run as far as i can and confirmed the the filenames are correct after this
I'm just coming online now. How can I help?
review here would be helpful ๐
aside from that, it should be good to merge main into llm again, and tag v0.17.0-llm.2 again
but i need to head out now, so not gonna be around
ok thanks! trying that now
@narrow juniper I tried to merge the latest main into llm but getting conflicts that look like this:
+++ b/cmd/dagger/shell.go
@@@ -11,6 -13,8 +11,11 @@@ import
"dagger.io/dagger"
"dagger.io/dagger/telemetry"
++<<<<<<< HEAD
++=======
+ "github.com/adrg/xdg"
+ "github.com/chzyer/readline"
++>>>>>>> main
"github.com/dagger/dagger/dagql/dagui"
"github.com/dagger/dagger/dagql/idtui"
"github.com/dagger/dagger/engine/client"
diff --cc dagql/idtui/frontend_pretty.go
index 160a049b2,773af294e..000000000
--- a/dagql/idtui/frontend_pretty.go
+++ b/dagql/idtui/frontend_pretty.go
@@@ -33,8 -29,6 +33,11 @@@ import
var isDark = termenv.HasDarkBackground()
++<<<<<<< HEAD
+var historyFile = filepath.Join(xdg.DataHome, "dagger", "histfile")
+
++=======
++>>>>>>> main
type frontendPretty struct {
dagui.FrontendOpts
Not sure if it's a "regular" merge conflict or related to your branch juggling
@cloud nacelle one of the PRs is merged now - so you can drop its merge during the rebase (tui-llm)
can take a look soon if you prefer
for those conflicts, you can just accept the top one (HEAD) for both
@narrow juniper there's also this one:
<<<<<<< HEAD
trace.WithAttributes(attribute.Bool(telemetry.UIRevealAttr, true)))
=======
trace.WithAttributes(attribute.Bool("dagger.io/ui.reveal", true)))
>>>>>>> main
span.End()
}()
func() {
ctx, span := Tracer().Start(ctx, "revealed message",
<<<<<<< HEAD
yes sorry vitztest/main.go
looks like there's another one cut off
hmm i can just do it if you want haha
not sure why that one's conflicting
Yeah if you don't mind taking the llm.2 release today, I'll switch to demo mode (need to get the demo to fit in 7mn for delivery this afternoon...)
(it doesn't have to be right this minute, I'm just anxious to get the latest improvement out to everyone, including anthropic support ๐
thank you ๐
force pushed (made sure to pull first)
- rebased all the merge commits out
- re-merged
shell-bbton top
will kick the tires a bit and then tag
Also a big thank you to @split bolt and @left onyx for dealing with this... The pain will be worth it!
tagged
I just installed it! Seems to work! 
Definitely! I am super happy to have pre-releases again. All the llm is many cherries on top of a pre-releases cake ๐
I am signing off for the day - it's been a super long week - but I trust that everything is OK?
Thank you @split bolt ๐
not sure what's up with these 3 failed jobs (also failed on rerun):
- https://github.com/dagger/dagger/actions/runs/13462770033/job/37622564007
- https://github.com/dagger/dagger/actions/runs/13462770033/job/37622564751
- https://github.com/dagger/dagger/actions/runs/13462770033/job/37622563204
but maybe we don't want these anyway
we at least don't expect go get dagger.io/dagger@v0.17.0-llm.2 to work (last failed job)
Looking
That is correct. We only need publish to work. The rest should be skipped on pre-release. #1342304973421150218 message
The checksums seem to be messed up for 0.17.0-llm.2:
โฏ curl -fsSL https://dl.dagger.io/dagger/install.sh | DAGGER_VERSION=0.17.0-llm.2 sh
sh debug downloading files into /var/folders/z5/n5y08g4j073_ld2d9019j3_80000gn/T/tmp.P5mncQsEgI
sh debug http_download https://dl.dagger.io/dagger/releases/0.17.0-llm.2/dagger_v0.17.0-llm.2_darwin_amd64.tar.gz
sh debug http_download https://dl.dagger.io/dagger/releases/0.17.0-llm.2/checksums.txt
sh err hash_sha256_verify unable to find checksum for '/var/folders/z5/n5y08g4j073_ld2d9019j3_80000gn/T/tmp.P5mncQsEgI/dagger_v0.17.0-llm.2_darwin_amd64.tar.gz' in '/var/folders/z5/n5y08g4j073_ld2d9019j3_80000gn/T/tmp.P5mncQsEgI/checksums.txt'
mmm my install went fine - maybe a race condition, it was temporarily working then broke ?
yeah a user just hit the missing checksum for amd64 too
some are there, some aren't - no idea why
linux amd64 is missing, which is a big one
I'm on Darwin/amd64
Maybe a race condition, because Darwin/amd64 is not working for me.
If I delete the files, want to tag & push again @narrow juniper?
I want to make sure that you can be self-sufficient with this release.
And I have a hard stop in 15 mins.
i'm not sure i understand the race condition 
hmm i was able to install on linux amd64 just now
I am going to invalidate Cloudfront for this path. What is the content of the file for you @narrow juniper ?
โฏ curl -v https://dl.dagger.io/dagger/releases/0.17.0-llm.2/checksums.txt
* Host dl.dagger.io:443 was resolved.
* IPv6: (none)
* IPv4: 18.154.84.92, 18.154.84.82, 18.154.84.126, 18.154.84.10
* Trying 18.154.84.92:443...
* Connected to dl.dagger.io (18.154.84.92) port 443
* ALPN: curl offers h2,http/1.1
* (304) (OUT), TLS handshake, Client hello (1):
* CAfile: /etc/ssl/cert.pem
* CApath: none
* (304) (IN), TLS handshake, Server hello (2):
* (304) (IN), TLS handshake, Unknown (8):
* (304) (IN), TLS handshake, Certificate (11):
* (304) (IN), TLS handshake, CERT verify (15):
* (304) (IN), TLS handshake, Finished (20):
* (304) (OUT), TLS handshake, Finished (20):
* SSL connection using TLSv1.3 / AEAD-AES128-GCM-SHA256 / [blank] / UNDEF
* ALPN: server accepted h2
* Server certificate:
* subject: CN=*.dagger.io
* start date: May 20 00:00:00 2024 GMT
* expire date: Jun 18 23:59:59 2025 GMT
* subjectAltName: host "dl.dagger.io" matched cert's "*.dagger.io"
* issuer: C=US; O=Amazon; CN=Amazon RSA 2048 M03
* SSL certificate verify ok.
* using HTTP/2
* [HTTP/2] [1] OPENED stream for https://dl.dagger.io/dagger/releases/0.17.0-llm.2/checksums.txt
* [HTTP/2] [1] [:method: GET]
* [HTTP/2] [1] [:scheme: https]
* [HTTP/2] [1] [:authority: dl.dagger.io]
* [HTTP/2] [1] [:path: /dagger/releases/0.17.0-llm.2/checksums.txt]
* [HTTP/2] [1] [user-agent: curl/8.7.1]
* [HTTP/2] [1] [accept: */*]
> GET /dagger/releases/0.17.0-llm.2/checksums.txt HTTP/2
> Host: dl.dagger.io
> User-Agent: curl/8.7.1
> Accept: */*
>
* Request completely sent off
< HTTP/2 200
< content-type: text/plain; charset=utf-8
< content-length: 839
< date: Fri, 21 Feb 2025 10:30:11 GMT
< last-modified: Fri, 21 Feb 2025 03:46:07 GMT
< etag: "75c3e4eedde04ca23bbb2e24738ba739"
< x-amz-server-side-encryption: AES256
< content-disposition: attachment;filename=checksums.txt
< accept-ranges: bytes
< server: AmazonS3
< x-cache: Hit from cloudfront
< via: 1.1 e457bcb869da0eddc75f07b0f42f8e3e.cloudfront.net (CloudFront)
< x-amz-cf-pop: LHR5-P7
< alt-svc: h3=":443"; ma=86400
< x-amz-cf-id: q4n_cYW5YIkbSvu_YRRAGMA4JaMmtNvLw-Bn12vWx5fkCeBz_qTw4Q==
< age: 29829
<
6a3e6f3b59bb661d8f774b0e1b1ad54bc16a991aff30df7987b15a72890d4f45 dagger_0.17.0-llm.2_darwin_amd64.tar.gz
b9291faccebccb39cc57de29efb5d19e24449e986e2e54c414089d9802410ea9 dagger_0.17.0-llm.2_darwin_arm64.tar.gz
9645815e084784036337a5e8298b8063cdd5b25dbfbd53398ebc595eb2891e6a dagger_0.17.0-llm.2_linux_amd64.tar.gz
54ad690d9f624e2056bfe3027d4764e7a604f1e6fd89a259bfab10398e52a966 dagger_0.17.0-llm.2_linux_arm64.tar.gz
859846e99bbb546761f450e8e51c002f77d050581190d000c0ae61e7ac2c043c dagger_0.17.0-llm.2_linux_armv7.tar.gz
d1e9920caa1e2704d9c929b21e66cfe4585f87a4be1d66b556cf0a0364e953ad dagger_0.17.0-llm.2_windows_amd64.zip
6ea5bb6b5ebbf2c6ee406f7b614c2d38079df63ce183df0b13f895f16e03b555 dagger_0.17.0-llm.2_windows_arm64.zip
ecaaec619f17fdbaf3ddc5687df8f7f186cc269f0b36b4a03db73bb54ed2c23a dagger_0.17.0-llm.2_windows_armv7.zip
* Connection #0 to host dl.dagger.io left intact
downloaded checksums.txt and i see this, which seems fine:
9396419036560ab58d5d1fbde949e5eaa0ac38502a5b7dd1b8e39d7365254ab8 dagger_v0.17.0-llm.2_darwin_amd64.tar.gz
9a91d5988db4327fc5e45fce65028f0f08f3eae780819b800a5e4a8d98dc7003 dagger_v0.17.0-llm.2_darwin_arm64.tar.gz
7be93ec4922b2dcbe8c45b8f2de74993204c8fbeca5670edef0a553c51349941 dagger_v0.17.0-llm.2_linux_amd64.tar.gz
51e6d27e8a64f2045c2ad877b0a784fe451dd1f57a5953b2932044284e4f6a3c dagger_v0.17.0-llm.2_linux_arm64.tar.gz
8c917506b291353d622a0cb94cdb7513477112cbfcfaf02501d2017cd0ab0e58 dagger_v0.17.0-llm.2_linux_armv7.tar.gz
8371a16e17a25c9e6164b57a4ada11b9b720be88384b2ac44d4dbf8ab773a8f6 dagger_v0.17.0-llm.2_windows_amd64.zip
6de4c43872ca751aba05525fb22d750b44f3040b693d034e9ef3b881a3350561 dagger_v0.17.0-llm.2_windows_arm64.zip
7a744651bcd0e282cccb68297b9c6246aca1ea8c4d7df9a28474f4f1a33d8696 dagger_v0.17.0-llm.2_windows_armv7.zip
Notice that mine was in the edge cache for 29829 secs.
I think that's the problem.
Different versions cached in Cloudfront.
Invalidating now.