Opened 19 months ago

Closed 2 weeks ago

#67295 closed defect (worksforme)

gexiv2 fails for valac cannot compile programs

Reported by: rmottola (Riccardo) Owned by: mascguy (Christopher Nielsen)
Priority: Normal Milestone:
Component: ports Version:
Keywords: leopard Cc:
Port: gexiv2

Description

Library m found: YES
Program /opt/local/bin/python3.10 (gi) found: YES (/opt/local/bin/python3.10) modules: gi
Checking if "BMFF support in exiv2" with dependency exiv2 compiles: YES 
Program g-ir-scanner found: YES (/opt/local/bin/g-ir-scanner)
Compiler for language vala for the build machine not found.

meson.build:42:19: ERROR: Vala compiler 'valac' can not compile programs

launched from the command line, valac executes with no files. This is Leopard i386

Attachments (1)

meson-log.txt (133.4 KB) - added by rmottola (Riccardo) 19 months ago.

Download all attachments as: .zip

Change History (9)

Changed 19 months ago by rmottola (Riccardo)

Attachment: meson-log.txt added

comment:1 Changed 19 months ago by jmroot (Joshua Root)

Keywords: leopard added
Owner: set to dbevans
Status: newassigned

comment:2 Changed 19 months ago by ryandesign (Ryan Carsten Schmidt)

Cc: mascguy added

The log says:

***MEMORY-ERROR***: valac[832]: GSlice: assertion failed: aligned_memory == (gpointer) addr

Searching for this error, I think GSlice is part of GLib. Which glib2 port is installed, and is it installed single-arch or universal?

port -vq installed name:^glib2

Back when glib2 used autotools there were extensive patches to ensure everything worked correctly for every arch, even when built universal. When the port was updated to a version that had dropped autotools support the patches had to be rewritten and it's possible some aspect was missed. Or it could be a new problem unrelated to that.

comment:3 Changed 19 months ago by kencu (Ken)

In a glib2-devel update some months ago, legacysupport was added to fix something erroring related to aligned memory access:

[04e909d5945557a659b2ab065bbbdd3e6447045f/macports-ports]

This then moved into glib2 on last week’s update.

Might be useful to see how the glib2 test suite liked the aligned memory access changes, assuming there are tests for that.

Last edited 19 months ago by ryandesign (Ryan Carsten Schmidt) (previous) (diff)

comment:4 Changed 19 months ago by kencu (Ken)

the assertion is coming from gslice.c it seems.

here’s a slightly older version found in debian’s source browser

https://sources.debian.org/src/glib2.0/2.48.0-1~bpo8+1/glib/gslice.c/#L1290

comment:5 Changed 18 months ago by mascguy (Christopher Nielsen)

Cc: mascguy removed
Owner: changed from dbevans to mascguy

comment:6 Changed 3 weeks ago by barracuda156

Is this still an issue on 10.5?

comment:7 Changed 2 weeks ago by rmottola (Riccardo)

I didn't see gexiv2 in my outdated packages at the last run, so I guess it compiled? I have a couple of continuously failing packages that do not allow a full upgrade to complete, but it is not one of them, so it seems solved.

comment:8 in reply to:  7 Changed 2 weeks ago by mascguy (Christopher Nielsen)

Resolution: worksforme
Status: assignedclosed

Replying to rmottola:

I didn't see gexiv2 in my outdated packages at the last run, so I guess it compiled? I have a couple of continuously failing packages that do not allow a full upgrade to complete, but it is not one of them, so it seems solved.

Sounds good, glad it's no longer an issue!

Note: See TracTickets for help on using tickets.