zlib-0.7.0.0

[alaendle]

zlib-0.7.0.0 (changelog) (Grandfathered dependencies) is out of bounds for:

• Agda-2.6.4.1 (>=0.6 && < 0.7). Andreas Abel [email protected] @andreasabel. Used by: library
• JuicyPixels-3.3.8 (>=0.5.3.1 && < 0.7). Vincent Berthoux [email protected] @Twinside. Used by: library
• cabal-install-3.10.2.1 (>=0.5.3 && < 0.7). Mikhail Glushenkov [email protected] @23Skidoo. Used by: library
• cborg-json-0.2.6.0 (>=0.5 && < 0.7). Ben Gamari [email protected] @bgamari. Used by: benchmark
• hackage-cli-0.1.0.1 (^>=0.6.1). Andreas Abel [email protected] @andreasabel. Used by: executable
• hackage-security-0.6.2.4 (>=0.5 && < 0.7). Andreas Abel [email protected] @andreasabel. Used by: library
• hyphenation-0.8.2 (>=0.5 && < 0.7). Edward Kmett [email protected] @ekmett. Used by: library
• io-streams-1.5.2.2 (>=0.5 && < 0.7). Gregory Collins [email protected] @gregorycollins. Used by: test-suite
• keter-2.1.3 (>=0.6.2 && < 0.7). Jappie Klooster [email protected] @jappeace. Used by: library
• pandoc-3.1.11.1 (>=0.5 && < 0.7). John MacFarlane [email protected] @jgm. Used by: library
• serialise-0.2.6.1 (>=0.5 && < 0.7). Grandfathered dependencies. Used by: benchmark
• snap-core-1.0.5.1 (>=0.5 && < 0.7). Grandfathered dependencies. Used by: test-suite
• wai-saml2-0.5 (>=0.6.0.0 && < 0.7). Michael B. Gale [email protected] @mbg. Used by: library

I'll add a constraint to zlib for now.

[andreasabel]

Fixed for Agda and hackage-cli.

[RyanGlScott]

I've applied a Hackage revision to hyphenation-0.8.2.

[andreasabel]

Fixed for hackage-security.

[jappeace]

fixed for keter by version bump

[jgm]

fixed in pandoc 3.1.12

[mbg]

Fixed for wai-saml2 with a metadata revision

[andreasabel]

JuicyPixels fixed by revision:

• Allow zlib-0.7; bump CI to GHCs 9.6.4 and 9.4.8 Twinside/Juicy.Pixels#223

[mihaimaruseac]

Current state is

zlib-0.7.0.0 (changelog) (Grandfathered dependencies) is out of bounds for:

• cborg-json-0.2.6.0 (>=0.5 && < 0.7). Ben Gamari [email protected] @bgamari. Used by: benchmark
• io-streams-1.5.2.2 (>=0.5 && < 0.7). Gregory Collins [email protected] @gregorycollins. Used by: test-suite
• serialise-0.2.6.1 (>=0.5 && < 0.7). Grandfathered dependencies. Used by: benchmark
• snap-core-1.0.5.1 (>=0.5 && < 0.7). Grandfathered dependencies. Used by: test-suite

I think we might just disable tests and benchmarks for these packages and enable zlib. Will give it a try later

[andreasabel]

io-streams fixed by revision

[andreasabel]

snap-core fixed by revision

[andreasabel]

Remaining serialise and cborg-json fixed by revision

[mihaimaruseac]

Thank you. This is now fixed.

[andreasabel]

Great @mihaimaruseac !

I see that is a lot of back-and-forth on master:
70f04a9 (HEAD -> master, origin/master) Temporarily allow back testing on aeson-schemas
117774d Remove zlib bound and fix #7309
f41163e Rvert
fb9a20b Try zlib agian
944dfac Revert to green
82c1389 Try zlib bound removal again
632a078 Add new package hs-captcha (#7374)
e113fd6 remove stale comments
b2c4199 Drop broken, stale package
547489a Expect test failure for aeson-schemas (#7376)
1283f46 Drop chronos-bench (#7375)
1f031a1 Restore to green
1f59606 Last test for upper bounds
fb3c7a7 Remove bound on scotty and fix #7345
e4bc5ef Yet one more try
55845bd Yet more try
0bc26d9 One more try again
1dca371 Try again
d351c2f Try agian
6867303 Remove bounds on ansi-terminal, fix #7306
629a7b9 Remove bounds on ansi-terminal

Is this a limitation of the tools? Otherwise, maybe one would only push successful updates to master...

[mihaimaruseac]

It's a mix of both. We can run the tooling locally, but for my system it takes too long. I switched on testing on the CI :)

For build errors, we need to test on the build system, it is too expensive to build everything locally

[andreasabel]

I switched on testing on the CI :)

When it comes to running check on CI, one could add some dedicated branches to the triggers, like ci-*:

stackage/.github/workflows/check.yml

Lines 1 to 6 in fe67452

	name: check
	on:
	push:
	branches: [ master ]
	pull_request:
	branches: [ master ]

I did so for my projects (before I switched to an "Always PR" policy for simplicity).

For build errors, we need to test on the build system,

I see. How does this work? Does the build system pull master from here?

[mpilgrem]

EDIT: I am going to re-raise the following as a separate issue, as I think pkg-config now being a de facto prerequisite to use Stackage with Stack (given the number of packages that depend on zlib) is a fundamental one.

I understand that Stack can only build zlib-0.7 with its default Cabal flags (specifically, pkg-config defaulting as true) if executable pkg-config is on the PATH. If zlib-0.7 is included in package sets without setting --flag zlib:-pkg-config should Stackage document that pkg-config is a prerequisite to use its package sets?

[andreasabel]

I am going to re-raise the following as a separate issue,

That would be:

• Is it intended that pkg-config be a prerequisite to use Stackage with Stack? #7387