From Alexey Yerin to ~sircmpwn/hare-dev
Signed-off-by: Alexey Yerin <yyp@disroot.org> --- cmd/hare/build/queue.ha | 2 +- cmd/hare/deps.ha | 2 +- cmd/haredoc/doc/sort.ha | 10 +++++----- cmd/haredoc/doc/util.ha | 2 +- glob/glob.ha | 6 +++--- hare/module/srcs.ha | 10 +++++----- hare/types/store.ha | 4 ++-- sort/+test.ha | 8 ++++---- sort/sort.ha | 13 +++++-------- time/date/virtual.ha | 4 ++-- 10 files changed, 29 insertions(+), 32 deletions(-) [message trimmed]
From Alexey Yerin to ~yerinalexey/public-inbox
Merged the patch series with a few modifications. Thanks a lot for working on this! I've made the repository parser ignore inline functions instead of parsing them (they are not useful for generating bindings) and merged module reclassification with other Makefile changes. To git.sr.ht:~yerinalexey/hare-gi 9e2c249..15f28a8 master -> master
From Alexey Yerin to ~yerinalexey/public-inbox
Applied. Thanks for the patch! To git.sr.ht:~yerinalexey/hare-gi d1d8d1a..9e2c249 master -> master
From Alexey Yerin to ~yerinalexey/public-inbox
Since all generated modules are now housed in gi::, it would be nice to remove the bindings/ directory, instead placing gi/ in the root of the repository. This allows to get rid of the HAREPATH fix for demos. Also, types::libc support module can be moved to gi::libc to be consistent with the rest. On Mon Mar 3, 2025 at 10:17 +0100, Drew DeVault wrote: > +MODULES_SHARED = types/libc > +MODULES_GI_SHARED = \ > + gi/glib \ > + gi/gobject \ > + gi/gio \ > + gi/gmodule \ > + gi/atk \
From Alexey Yerin to ~mpu/qbe
Instruction encoding for shift-with-immediate usually has more
restrictions compared to other immediate instructions. On amd64, the
maximum shift amount is 255; on rv64 it is 31 or 63, depending on the
result size. Arm64 isn't affected.
QBE used to just pass large amount values to the assembler which would
then fail because it couldn't encode them. Now the values are taken
modulo result size prior to emitting, which should avoid the issue.
Constant shift amounts are generally never larger than result size so
the assembler failure is unlikely arise in a real codebase. I only ran
into it by accidentally using the wrong constant for a left shift.
---
Not sure whether simpl.c is the right place to put this fix. It seemed
[message trimmed]
From Alexey Yerin to ~yerinalexey/public-inbox
On Fri Feb 21, 2025 at 11:49 +0100, Drew DeVault wrote: > -HAREPATH_CMD = hare version -v | awk 'harepath { gsub(/^\t/, ""); print }; /^HAREPATH/ { harepath = 1 }' | tr '\n' ':' > +HAREPATH_CMD = hare version -vv | grep '^HAREPATH' | cut -d'=' -f2- This doesn't work correctly if cmd/hare decides to quote the value of HAREPATH. On my system it outputs `HAREPATH='<...>'` which can't be used without unquoting first. I pushed an improved implementation that unquotes the value of HAREPATH, and only then adds the bindings directory to it.
From Alexey Yerin to ~yerinalexey/public-inbox
Applied, thanks! To git.sr.ht:~yerinalexey/hare-gi a53329a..27daf65 master -> master
From Alexey Yerin to ~sircmpwn/hare-dev
When truncated module paths are used in debug info, debuggers can't find the correct source file. A similar issue can occur when printing harec errors with line context. Fixes: https://todo.sr.ht/~sircmpwn/hare/934 Signed-off-by: Alexey Yerin <yyp@disroot.org> --- include/util.h | 2 ++ src/emit.c | 2 +- src/main.c | 3 +++ src/util.c | 3 ++- 4 files changed, 8 insertions(+), 2 deletions(-) diff --git a/include/util.h b/include/util.h [message trimmed]
From Alexey Yerin to ~sircmpwn/hare-dev
--- pixbuf/buffer.ha | 4 ++-- pixbuf/convert.ha | 4 ++-- pixbuf/test.ha | 16 ++++++++-------- 3 files changed, 12 insertions(+), 12 deletions(-) diff --git a/pixbuf/buffer.ha b/pixbuf/buffer.ha index 13af734..8a5b25e 100644 --- a/pixbuf/buffer.ha +++ b/pixbuf/buffer.ha @@ -16,10 +16,10 @@ export fn newbuffer( width: size, height: size, stride: size, [message trimmed]
From Alexey Yerin to ~yerinalexey/public-inbox
Thanks for the patch! To git.sr.ht:~yerinalexey/hare-gi e5d712e..a53329a master -> master