You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
i was hoping to try out the 'mars-gtk' experiment; but pharo is not behaving well on my system
the behavior is the same with 80/headless-latest, 90/headless-latest, 90/headless-stable, and 90/stable - upon launching pharo, i get this error (referring to libgit2):
$ ./pharo64-linux-headless-latest/pharo ./pharo64/Pharo9.0-SNAPSHOT-64bit-8ce6236.image
External module not found
ExternalLibraryFunction(Object)>>error:
ExternalLibraryFunction(Object)>>externalCallFailed
ExternalLibraryFunction(ExternalFunction)>>invokeWithArguments:
LGitLibrary>>libgit2_init
FFICalloutAPI>>function:library:
LGitLibrary(Object)>>ffiCall:library:options:
LGitLibrary(Object)>>ffiCall:library:
LGitLibrary(Object)>>ffiCall:
LGitLibrary>>libgit2_init
the library file looks fine
$ file lib/libgit2.so.0.25.1
lib/libgit2.so.0.25.1: ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), statically linked, BuildID[sha1]=bca31b8ef88a657faffdaa55c5960250ecbc7b9a, with debug_info, not stripped
i "fixed" that with the "latest" VM, by deleting the libgit2 libraries from the lib/ dir and adding a symlink to the system libgit2 lib - by "fixed", i mean that it allowed pharo to start without the FFI error - this trick worked for all image/VM combinations that i tried
however, upon loading 'mars-gtk' with metacello, the "latest" VM crashed with a segfault; so its not clear that my "fix" was an improvement
after some poking around in the image, i found that metacello was trying to use SSH to download 'mars-gtk' and could not find any credentials - i changed the setting of "iceberg" in its GUI to prefer HTTPS and that allowed the git clone to start - it chugged away for about a minute looking quite normal, presenting several progress-bar widgets to completion - then it eventually crashed with another segfault - the problem seems to be with libssl now - perhaps related to #70 (closed as not-a-bug)
with the stable VM, replacing libgit2 as i described previously, eliminates that initial error also, as with the "latest" VM; but loading a package with metacello, fails immediately with this error, apparently before SSL becomes relevant:
Instance of Array did not understand #(#SqueakFFIPrims #primitiveCalloutWithArgs 4 13)
Array(Object)>>doesNotUnderstand: #(#SqueakFFIPrims #primitiveCalloutWithArgs 4 13)
Array(ExternalFunction)>>invokeWithArguments:
LGitRemote>>remote_fetch:refspecs:opts:reflog_message:
LGitSafeFFICalloutAPI(FFICalloutAPI)>>function:library:
LGitSafeFFICalloutAPI(FFICalloutAPI)>>function:module:
LGitRemote(LGitExternalObject)>>ffiCallSafely:options:
LGitRemote>>remote_fetch:refspecs:opts:reflog_message:
the call stack for "Array(ExternalFunction)>>invokeWithArguments:" show that 'nil' was passed in
maybe it would help if i compiled the VM from source on my system - would the 'headless' branch of this repo be compatible with the 'mars-gtk' package?
The text was updated successfully, but these errors were encountered:
i was hoping to try out the 'mars-gtk' experiment; but pharo is not behaving well on my system
the behavior is the same with 80/headless-latest, 90/headless-latest, 90/headless-stable, and 90/stable - upon launching pharo, i get this error (referring to libgit2):
the library file looks fine
i "fixed" that with the "latest" VM, by deleting the libgit2 libraries from the lib/ dir and adding a symlink to the system libgit2 lib - by "fixed", i mean that it allowed pharo to start without the FFI error - this trick worked for all image/VM combinations that i tried
however, upon loading 'mars-gtk' with metacello, the "latest" VM crashed with a segfault; so its not clear that my "fix" was an improvement
after some poking around in the image, i found that metacello was trying to use SSH to download 'mars-gtk' and could not find any credentials - i changed the setting of "iceberg" in its GUI to prefer HTTPS and that allowed the git clone to start - it chugged away for about a minute looking quite normal, presenting several progress-bar widgets to completion - then it eventually crashed with another segfault - the problem seems to be with libssl now - perhaps related to #70 (closed as not-a-bug)
with the stable VM, replacing libgit2 as i described previously, eliminates that initial error also, as with the "latest" VM; but loading a package with metacello, fails immediately with this error, apparently before SSL becomes relevant:
the call stack for "Array(ExternalFunction)>>invokeWithArguments:" show that 'nil' was passed in
maybe it would help if i compiled the VM from source on my system - would the 'headless' branch of this repo be compatible with the 'mars-gtk' package?
The text was updated successfully, but these errors were encountered: