Toots for ptesarik@fosstodon.org account

Written by Petr Tesarik on 2025-01-23 at 13:28

Hey, @vbabka I believe Jack deleted more kernel lines in 6.13 than you ever have. What will you kill next to make up the difference?

https://lwn.net/Articles/1004998/

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2025-01-18 at 16:04

Just received a postcard from @conservancy. Thank you for your work!

=> View attached media

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2025-01-15 at 11:18

Another real-world lesson on why optimizing for #benchmark is bad:

https://github.com/evenfurther/pathfinding/issues/639

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2025-01-07 at 12:56

If your patch has 3 lines of context and does not remove any lines, it will always apply with --fuzz=3.

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2025-01-02 at 12:57

I have just reviewed https://github.com/ptesarik/libkdumpfile/issues/43

At this point, I believe there must be a decent way to generate HTML documentation and man pages from comments in source code, it's just that Doxygen is not up to the job.

Any suggestions? Can Sphinx do it out of the box (or with some minimal hacking)?

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2024-12-30 at 14:35

Remember: Every line of code you write is a liability.

Hm, and is every line of code I delete an asset?

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2024-12-30 at 07:58

A use-after-free bug found thanks to @obshq 👍

Tests were failing only with Python 3.12 (no older nor newer versions) and only for a couple of architectures (both 32-bit), but the bug was always there, just masked by the Python garbage collector. Thankfully, OBS automatically builds my package for a wide variety of targets, revealing the bug in no time.

Anyway, if you use pykdumpfile, upgrade to https://pypi.org/project/pykdumpfile/0.5.5.1/

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2024-12-26 at 20:36

I've just deleted 10k LOC. It's not in the #Linux kernel, but I'm still proud of myself.

https://github.com/ptesarik/libkdumpfile/commit/40c9ff981f1f3bd968af37a50b50c3478d8267cd

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2024-12-25 at 20:48

Christmas release of libkdumpfile is here. This is the LAST release which still includes the legacy Python bindings. You should really switch to pykdumpfile now.

https://github.com/ptesarik/libkdumpfile/releases/tag/v0.5.5

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2024-12-19 at 15:35

I hate libopcodes from binutils for requiring me to add an unknown set of additional libraries to the linker command line.

I also hate GNU autotools for making it extremely difficult to write the automated checks.

But nothing can stop a determined developer, even if it means writing a hundred lines in M4.

And that's why the world can see

https://github.com/ptesarik/libkdumpfile/commit/26282fcb66488ed40b0222e0b870e14101053ec7

Suggested by: @brenns10

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2024-12-12 at 10:29

Just discovered config DRM_I915_SELFTEST_BROKEN.

…

Recommended for masochistic driver developers only.

It defaults to N, which is fine, but the help text could be changed to: If in doubt, say "Y".

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2024-12-09 at 12:44

Has anyone tried to improve #Linux TLB flushing so that a single INVLPG is used to invalidate a huge page?

Documentation says: “Despite the fact that a single individual flush on x86 is guaranteed to flush a full 2MB, hugetlbfs always uses the full flushes. THP is treated exactly the same as normal memory.”

https://www.kernel.org/doc/html/latest/arch/x86/tlb.html

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2024-12-02 at 18:04

@vbabka Turns out I was too quick and forgot how I wanted to fix attribute references in libkdumpfile. So, this PR is retracted for now.

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2024-12-02 at 16:41

@vbabka This is not the kernel, so it does not count for real, but I'm still proud of myself for the negative diff.

https://github.com/osandov/drgn/pull/451/files

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2024-11-29 at 19:51

Now, seeing this toot got a few favourites, who of you guys implemented that feature? 😉

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2024-11-29 at 15:29

@brenns10 Finally fixed this long-standing issue:

https://github.com/ptesarik/libkdumpfile/issues/81

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2024-11-29 at 15:14

Just discovered a cool #magit feature: When you write a new commit and mention an issue using the usual hash-notation, it immediately becomes a hyperlink to that issue on GitHub.

Wow! #Emacs

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2024-11-29 at 09:02

Over the last 7 days I've made some changes to libkdumpfile which make it work much better with QEMU ELF dumps. Cc @brenns10

https://github.com/ptesarik/libkdumpfile/blob/c5301ced60ad516a4eb5e72648cfd7147a6be831/NEWS

EDIT: I feel I should fix a few long-standing bugs and make a new release before Christmas.

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2024-11-19 at 12:24

OK, I believe that testing of const data could be improved a bit.

My first #hackweek24 patch:

https://lore.kernel.org/linux-mm/cover.1732016064.git.ptesarik@suse.com/T/

=> More informations about this toot | View the thread

Written by Petr Tesarik on 2024-11-18 at 10:37

@vbabka @ljs Hey, mm folks, I'm trying to kill those pesky DMA* zones… Is there really a single per-pgdat LRU list, so the reclaim code always wades through all folios before it finds those few missing DMA-zone pages at the end of the list?

It's not like that, is it?

…

Is it?

=> More informations about this toot | View the thread

=> This profile with reblog | Go to ptesarik@fosstodon.org account