Попробуйте доставить ghc-devel

Хотя, это конечно нелогично. Нужно будет поправить.

Доустановка ghc-devel помогает.

ghc-9.6.6-alt2 -> sisyphus:

 Thu Apr 24 2025 Leonid Znamenok <respublica@altlinux> 9.6.6-alt2
 - Rebuild with dependency on ghc-devel (Closes: #53978)

$ cat ~/hasher/aptbox/etc/apt/sources.list
rpm-dir file:/tmp/.private/imz/hasher/repo x86_64 hasher
rpm [alt] file:/ALT/Sisyphus x86_64 classic debuginfo checkinstall
rpm [alt] file:/ALT/Sisyphus noarch classic checkinstall
rpm [alt] file:/ALT/Sisyphus x86_64-i586 classic
$ hsh-install --workdir ~/hasher/ ghc-ghci
<13>Feb 13 12:37:30 rpmi: ghc9.6-ghc-prim-1:0.10.0-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-ghc-bignum-1:1.3-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-base-1:4.18.3.0-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-array-1:0.5.8.0-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-deepseq-1:1.4.8.1-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-pretty-1:1.1.3.6-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-time-1:1.12.2-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-stm-1:2.5.1.0-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-ghc-boot-th-1:9.6.7-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-template-haskell-1:2.20.0.0-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-bytestring-1:0.11.5.4-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-containers-1:0.6.7-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-binary-1:0.8.9.1-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-ghc-heap-1:9.6.7-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-transformers-1:0.6.1.0-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-mtl-1:2.3.1-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-exceptions-1:0.10.7-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-filepath-1:1.4.301.0-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-unix-1:2.8.6.0-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-directory-1:1.3.8.5-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-ghc-boot-1:9.6.7-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc9.6-ghci-1:9.6.7-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:37:30 rpmi: ghc-ghci-1:9.6.7-alt1 sisyphus+383971.200.3.3 1747159306 installed
$ hsh-shell --mount=/proc,/dev/pts --workdir ~/hasher/ 
[builder@localhost .in]$ ghci
-bash: ghci: command not found
[builder@localhost .in]$

$ hsh-install --workdir ~/hasher/ ghc9.6-common
<13>Feb 13 12:40:19 rpmi: ghc9.6-process-1:1.6.19.0-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:40:19 rpmi: rpm-macros-ghc9.6-common-1:9.6.7-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:40:19 rpmi: ghc9.6-hpc-1:0.6.2.0-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:40:20 rpmi: ghc9.6-ghc-1:9.6.7-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:40:20 rpmi: ghc9.6-xhtml-1:3000.2.2.1-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:40:20 rpmi: ghc9.6-libiserv-1:9.6.7-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:40:20 rpmi: ghc9.6-terminfo-1:0.4.1.6-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:40:20 rpmi: ghc9.6-haskeline-1:0.8.2.1-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:40:20 rpmi: libgmp-devel-6.3.0-alt1 sisyphus+328118.100.2.1 1693294124 installed
<13>Feb 13 12:40:20 rpmi: libffi-devel-1:3.5.2-alt1 sisyphus+397156.100.1.1 1760464522 installed
<13>Feb 13 12:40:20 rpmi: haskell-filetrigger-0.1.0-alt1 sisyphus+378577.100.1.1 1742391763 installed
<13>Feb 13 12:40:20 rpmi: ghc9.6-text-1:2.0.2-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:40:20 rpmi: ghc9.6-parsec-1:3.1.16.1-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:40:21 rpmi: ghc9.6-Cabal-syntax-1:3.10.3.0-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:40:21 rpmi: ghc9.6-Cabal-1:3.10.3.0-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:40:23 rpmi: ghc9.6-1:9.6.7-alt1 sisyphus+384137.100.1.1 1747235447 installed
<13>Feb 13 12:40:23 rpmi: ghc9.6-common-1:9.6.7-alt1 sisyphus+384137.100.1.1 1747235447 installed
$ hsh-run --mount=/proc,/dev/pts --workdir ~/hasher/ -- ghci; echo $?
GHCi, version 9.6.7: https://www.haskell.org/ghc/  :? for help

<interactive>:1:1: error: [GHC-52243]
    Failed to load interface for `GHC.Types'
    no unit id matching `ghc-prim' was found
1
$ hsh-run --mount=/proc,/dev/pts --workdir ~/hasher/ -- rpm -qa 'ghc*-devel'; echo $?
0

ghc-devel is not installed, contrary to that fix:


https://bugzilla.altlinux.org/show_bug.cgi?id=57857#c1 :

> A related fix that didn't help:
> 
> ==ghc-ghci==
> * Thu Apr 24 2025 Leonid Znamenok <respublica@altlinux.org> 9.6.6-alt2
> - Rebuild with dependency on ghc-devel (Closes: #53978)
> 
> https://bugzilla.altlinux.org/53978


But there is still no dependency on ghc-devel:

apt> depends ghc-ghci
ghc-ghci-9.6.6-alt2:p11+387571.100.1.1@1750762280
  Depends: ghc9.6-ghci
apt> depends ghc9.6-ghci
ghc9.6-ghci-9.6.6-alt3:p11+383990.200.3.1@1750711102
  Depends: ghc9.6-ghc-boot = 9.6.6-alt3:p11+383990.200.3.1
  Depends: ghc9.6-ghc-heap = 9.6.6-alt3:p11+383990.200.3.1
apt> 


BTW, installing ghc-devel would mean consuming a lot of additional space on the filesystem. Is it necessary for ghci to work?

apt> install ghc-devel
Unrequested changes are needed to execute this operation.
The following NEW packages will be installed:
  ghc                        ghc9.6-base-devel        ghc9.6-devel             ghc9.6-ghc-boot-devel     ghc9.6-ghc-heap-devel   ghc9.6-integer-gmp        ghc9.6-pretty-devel            ghc9.6-text-devel          ghc9.6-unix-devel
  ghc-devel                  ghc9.6-binary-devel      ghc9.6-directory-devel   ghc9.6-ghc-boot-th-devel  ghc9.6-ghc-prim-devel   ghc9.6-integer-gmp-devel  ghc9.6-process-devel           ghc9.6-time-devel          ghc9.6-xhtml-devel
  ghc9.6-Cabal-devel         ghc9.6-bytestring-devel  ghc9.6-exceptions-devel  ghc9.6-ghc-compact        ghc9.6-ghci-devel       ghc9.6-libiserv-devel     ghc9.6-stm-devel               ghc9.6-transformers-devel
  ghc9.6-Cabal-syntax-devel  ghc9.6-containers-devel  ghc9.6-filepath-devel    ghc9.6-ghc-compact-devel  ghc9.6-haskeline-devel  ghc9.6-mtl-devel          ghc9.6-template-haskell-devel  ghc9.6-ucd2haskell
  ghc9.6-array-devel         ghc9.6-deepseq-devel     ghc9.6-ghc-bignum-devel  ghc9.6-ghc-devel          ghc9.6-hpc-devel        ghc9.6-parsec-devel       ghc9.6-terminfo-devel          ghc9.6-ucd2haskell-devel
0 upgraded, 42 newly installed, 0 removed and 0 kept.
Will need more 271MB of archives.
After unpacking will need more 3349MB of disk space.
Do you want to continue? [Y/n] n

(Ответ для Ivan Zakharyaschev на комментарий #4)
> ghc-devel is not installed, contrary to that fix:
> https://bugzilla.altlinux.org/show_bug.cgi?id=57857#c1 :
> ...
> But there is still no dependency on ghc-devel

At the moment, GHC is packaged in such a way that ghc9.6-ghci does not contain an executable file (/usr/bin/ghci9.6.7), but only the GHCi module built as a dynamic library:

https://hackage.haskell.org/package/ghci

$ rpm -ql ghc9.6-ghci
/usr/lib64/ghc-9.6.7/lib/x86_64-linux-ghc-9.6.7/libHSghci-9.6.7-ghc9.6.7.so


A library is an independent entity and can be installed (and used) separately from the interpreter itself. This possibility is more theoretical than practical. There are several modules that use GHCi without depending on the main GHC module, but in ALT they are not represented in any way.

Moreover, if we look at what GHCi actually is:

$ cat $(which ghci-9.6.7)
#!/bin/sh
exedir="/usr/lib64/ghc-9.6.7/bin"
exeprog="./ghci-9.6.7"
executablename="/usr/lib64/ghc-9.6.7/bin/./ghci-9.6.7"
bindir="/usr/bin"
libdir="/usr/lib64/ghc-9.6.7/lib"
docdir="/usr/share/doc/ghc-9.6.7"
includedir="/usr/include"

executable="$bindir/ghc-9.6.7"
exec $executable --interactive "$@"


it becomes clear that separating the interpreter into a package that has mininal dependencies and which is independent from the compiler is impossible. That is, GHC itself must always be installed together with GHCi, and GHC in turn requires the full set of -devel libraries.

In any case, answering the question:
> BTW, installing ghc-devel would mean consuming a lot of additional space on the filesystem. Is it necessary for ghci to work?

Yes, these libraries are mandatory.

-----


Reading this again now, I realize that these considerations are not the most obvious ones. In principle, I could move GHCi (the script shown above) into the ghc(9.6)-ghci package. That would make the internal structure a bit more obvious, but at the same time it would give it the full set of dependencies on all bundled GHC libraries.

Thank you for the explanation!

Maybe it's not a bug then.

(Ответ для Leonid Znamenok на комментарий #5)
> (Ответ для Ivan Zakharyaschev на комментарий #4)
> > ghc-devel is not installed, contrary to that fix:
> > https://bugzilla.altlinux.org/show_bug.cgi?id=57857#c1 :
> > ...
> > But there is still no dependency on ghc-devel
> 
> At the moment, GHC is packaged in such a way that ghc9.6-ghci does not
> contain an executable file (/usr/bin/ghci9.6.7), but only the GHCi module
> built as a dynamic library:

Maybe the problem was that I didn't realize this from the Summary of ghc-ghci:

Summary     : Haskell ghci runtime files

$ rpm -q ghc{9.6,}-ghci --qf='%{SUMMARY}\n'
Haskell ghci runtime files
Haskell ghci runtime files


I thought that if the executable /usr/bin/ghci is in ghc9.6-common, then the missing part for using ghci must have been "Haskell ghci runtime files". (But I wanted to outsmart the set of packages -- not wanting to install the gigabytes of *-devel packages.)

> In any case, answering the question:
> > BTW, installing ghc-devel would mean consuming a lot of additional space on the filesystem. Is it necessary for ghci to work?
> 
> Yes, these libraries are mandatory.

I see. Thank you for the explanation!

> 
> -----
> 
> 
> Reading this again now, I realize that these considerations are not the most
> obvious ones. In principle, I could move GHCi (the script shown above) into
> the ghc(9.6)-ghci package. That would make the internal structure a bit more
> obvious, but at the same time it would give it the full set of dependencies
> on all bundled GHC libraries.

I just don't see the purpose of ghc-ghci package in the current situation. Its name is a bit misleading. After your explanations, I'd be inclined to leave ghc9.6-ghci module with minimal dependencies (for the users of this module as a library). Now I see the "purpose": Then ghc-ghci logically is just the "standard" version of it in Sisyphus. But the name is a but misleading, which made me expect that it will make ghci work...

Maybe just add a note to the Summary of the package ghc{9.6,}-ghci, and maybe make another package (ghci) if installing ghc is not enough for it to work...

(Ответ для Ivan Zakharyaschev на комментарий #6)

> Maybe just add a note to the Summary of the package ghc{9.6,}-ghci, and

Not true:

> maybe make another package (ghci) if installing ghc is not enough for it to
> work...

My actual story and an inconsistency I noticed:

if I install ghc9.6 and ghc9.6-common first, ghci doesn't work. If I install ghc9.6-ghci then, ghci still doesn't work. And even if I install ghc-ghci, ghci still doesn't work. 

They don't have a dependency on ghc9.6-devel (unlike the ghc package):

$ rpm -q ghc9.6 -R
libffi-devel
libgmp-devel
glibc-gconv-modules
haskell-filetrigger
/bin/sh
/lib64/ld-linux-x86-64.so.2
ghc9.6-Cabal = 3.10.3.0-alt3:p11+383990.200.3.1
ghc9.6-haskeline = 0.8.2.1-alt3:p11+383990.200.3.1
ghc9.6-libiserv = 9.6.6-alt3:p11+383990.200.3.1
ghc9.6-xhtml = 3000.2.2.1-alt3:p11+383990.200.3.1
ghc9.6-ghc = 9.6.6-alt3:p11+383990.200.3.1
rpmlib(PayloadIsXz)
$ rpm -q ghc -R
ghc9.6 = 9.6.6
ghc9.6-common = 9.6.6
ghc-devel = 9.6.6-alt2:p11+387571.100.1.1
rpmlib(PayloadIsLzma)
$

I believe there are reasons for a minimal ghc9.6 and it's ok, but I think that the difference between ghc9.6 and ghc might be misleading, it leads to thinking that ghc9.6 is also enough, but it's not enough, for example, for compiling simple programs (needing Prelude):

$ hsh-run --mount=/proc,/dev/pts --workdir ~/hasher/ -- sh -c 'cd && cat >hello.hs && ghc hello.hs && ./hello' <<<'main = putStrLn "hello"'; echo $?
[1 of 2] Compiling Main             ( hello.hs, hello.o )

hello.hs:1:1: error:
    Could not find module `Prelude'
    Use -v (or `:set -v` in ghci) to see a list of the files searched for.
  |
1 | main = putStrLn "hello"
  | ^
1

After installing ghc9.6-devel it compiles, and ghci works.

And in a clean hasher, installing just only ghc9.6-devel makes it compile common programs like hello.hs and makes ghci work.

So, just ghc9.6-devel is the right package to install (for compiling common programs and ghci). Or ghc-devel.


But this wasn't quite clear. I don't know what to suggest. Maybe nothing can be improved here.

Another thing I've just noticed: There is no practical difference between ghc-devel and ghc:

First, I install ghc-devel. Then installing ghc adds no packages at all. Having two packages doesn't make sense from this point of view.

hsh --ini ~/hasher && hsh-install --workdir ~/hasher/ ghc-devel
hsh-install ~/hasher/ ghc

Most likely, things can indeed be simplified here by removing ghc(9.6)-devel altogether.
GHC requires all ghc-*-devel packages to function properly anyway. So having two separate entities looks redundant and really misleading.

I’ll take a look at what can be done here.

Thanks for the contribution!

(Ответ для Ivan Zakharyaschev на комментарий #8)
> My actual story and an inconsistency I noticed:
> 
> if I install ghc9.6 and ghc9.6-common first, ghci doesn't work. If I install
> ghc9.6-ghci then, ghci still doesn't work. And even if I install ghc-ghci,
> ghci still doesn't work. 

Ok, ghc-ghci was a wrong idea.

The actually working way for a minimal working installation of ghci (and runghc) is adding:

apt> install ghc9.6-ghc-prim-devel 
Unrequested changes are needed to execute this operation.
The following NEW packages will be installed:
  ghc9.6-base-devel  ghc9.6-ghc-bignum-devel  ghc9.6-ghc-prim-devel


Before (with just ghc9.6 and ghc9.6-common):

[imz@cheval ~]$ runghc <<<'main = print (4/3)'

<interactive>:1:1: error: [GHC-52243]
    Failed to load interface for ‘GHC.Types’
    no unit id matching ‘ghc-prim’ was found
[imz@cheval ~]$ ghci
GHCi, version 9.6.6: https://www.haskell.org/ghc/  :? for help

<interactive>:1:1: error: [GHC-52243]
    Failed to load interface for ‘GHC.Types’
    no unit id matching ‘ghc-prim’ was found

After (having installed ghc9.6-ghc-prim-devel):

[imz@cheval ~]$ runghc <<<'main = print (4/3)'
1.3333333333333333
[imz@cheval ~]$ ghci
GHCi, version 9.6.6: https://www.haskell.org/ghc/  :? for help
ghci> 4/3
1.3333333333333333
ghci> 
Leaving GHCi.

This saves space compared to the recommended installation of the bigger ghc-devel.

Yes, even the ghc9.6-base-devel package alone is sufficient for it to work.
I am currently preparing a task with dependency reduce and hope that it'll be in Sisyphus soon.

Additionally, I think that it's worth adding the infix "-package-" to the names of library packages. This is the standard naming for distributed libraries in GHC (hackage.haskell.org/packages/) and probably will help avoid such misunderstandings, as happened with the ghci package upon. Package name with a "-package-" infix sounds a bit redundant, but cleaner than "ghc-SMTHG".

(Ответ для Leonid Znamenok на комментарий #12)
> Yes, even the ghc9.6-base-devel package alone is sufficient for it to work.
> I am currently preparing a task with dependency reduce and hope that it'll
> be in Sisyphus soon.

They are interdependent:

[imz@team ~]$ cat ~/hasher/aptbox/etc/apt/sources.list
#rpm-dir file:/home/imz/repo x86_64 hasher
rpm [alt] file:///ALT Sisyphus/x86_64 classic
rpm [alt] file:///ALT Sisyphus/noarch classic
[imz@team ~]$ ~/hasher/aptbox/apt-cache  depends ghc9.6-ghc-prim-devel
ghc9.6-ghc-prim-devel-1:0.10.0-alt1:sisyphus+384137.100.1.1@1747235447
  Depends: ghc9.6 = 1:9.6.7-alt1:sisyphus+384137.100.1.1
  Depends: ghc9.6-base-devel = 1:4.18.3.0-alt1:sisyphus+384137.100.1.1
[imz@team ~]$ ~/hasher/aptbox/apt-cache whatdepends ghc9.6-ghc-prim-devel | grep -Fe base -A3
  ghc9.6-base-devel-1:4.18.3.0-alt1:sisyphus+384137.100.1.1@1747235447
    Depends: ghc9.6-ghc-prim-devel = 1:0.10.0-alt1:sisyphus+384137.100.1.1

(Ответ для Leonid Znamenok на комментарий #12)

> Additionally, I think that it's worth adding the infix "-package-" to the
> names of library packages. This is the standard naming for distributed
> libraries in GHC (hackage.haskell.org/packages/) and probably will help
> avoid such misunderstandings, as happened with the ghci package upon.
> Package name with a "-package-" infix sounds a bit redundant, but cleaner
> than "ghc-SMTHG".

This sounds reasonable.

(Ответ для Leonid Znamenok на комментарий #12)

> Additionally, I think that it's worth adding the infix "-package-" to the
> names of library packages. This is the standard naming for distributed
> libraries in GHC (hackage.haskell.org/packages/) and probably will help
> avoid such misunderstandings, as happened with the ghci package upon.
> Package name with a "-package-" infix sounds a bit redundant, but cleaner
> than "ghc-SMTHG".

This sounds reasonable.

(Ответ для Ivan Zakharyaschev на комментарий #13)

> They are interdependent:

> [imz@team ~]$ ~/hasher/aptbox/apt-cache  depends ghc9.6-ghc-prim-devel
> ghc9.6-ghc-prim-devel-1:0.10.0-alt1:sisyphus+384137.100.1.1@1747235447
>   Depends: ghc9.6 = 1:9.6.7-alt1:sisyphus+384137.100.1.1
>   Depends: ghc9.6-base-devel = 1:4.18.3.0-alt1:sisyphus+384137.100.1.1

Officially, there is no such dependency:

https://hackage.haskell.org/package/ghc-prim-0.10.0/dependencies

Dependencies for ghc-prim-0.10.0

Libraries

ghc-prim

* rts (>=1.0 && <1.1)

No "base". This could be important for uses where people don't need the much bigger package "base", but I don't know whether such situation is practically real.

(Ответ для Ivan Zakharyaschev на комментарий #16)
> (Ответ для Ivan Zakharyaschev на комментарий #13)
> 
> > They are interdependent:
> 
> > [imz@team ~]$ ~/hasher/aptbox/apt-cache  depends ghc9.6-ghc-prim-devel
> > ghc9.6-ghc-prim-devel-1:0.10.0-alt1:sisyphus+384137.100.1.1@1747235447
> >   Depends: ghc9.6 = 1:9.6.7-alt1:sisyphus+384137.100.1.1
> >   Depends: ghc9.6-base-devel = 1:4.18.3.0-alt1:sisyphus+384137.100.1.1
> 
> Officially, there is no such dependency:
> 
> https://hackage.haskell.org/package/ghc-prim-0.10.0/dependencies
> 
> Dependencies for ghc-prim-0.10.0
> 
> Libraries
> 
> ghc-prim
> 
> * rts (>=1.0 && <1.1)
> 
> No "base". This could be important for uses where people don't need the much
> bigger package "base", but I don't know whether such situation is
> practically real.

The rts (Runtime System) is a very special package in the GHC ecosystem. It consists of C code (not Haskell) and contains the most fundamental components required for program execution (GC, evaluation, scheduler, exceptions, etc.).

At the same time it is:
a) Inseparable from any code compiled with GHC and required for its build
b) De facto not a standalone package (the page https://hackage.haskell.org/package/rts is unavailable and it cannot be downloaded externally; it is obtained only as a part of the compiler distribution)
c) Appears as a dependency only for two internal and fundamental GHC packages: base and ghc-prim

For these reasons (primarily reason b), rts has been moved into the base package. In practice base is required to compile almost any Haskell project anyway. There might exist some degenerate cases where base is not used, but I am not aware of any.

-----

Although now I think that from a cleaner architectural point of view it would make more sense to move rts into ghc-prim. Perhaps this could also help break the interdependency between base and ghc-prim.

(Ответ для Leonid Znamenok на комментарий #17)

Thanks for the details!

> -----
> 
> Although now I think that from a cleaner architectural point of view it
> would make more sense to move rts into ghc-prim. Perhaps this could also
> help break the interdependency between base and ghc-prim.

I agree this is a good idea.