Well, that's reassuring, at least -- firstly, there's a mere 82030 lines of unified diff between those versions (https://github.com/curl/curl/compare/curl-8_11_1...curl-8_12_0) and secondly, the upstream changes that fixed the configure script (amongst other amigaos3 changes) went into 8.12.0, so that was already the likeliest culprit.
J.
Right, it appears that 8.11.1 is OK, but 8.12.0 is not.
On 21/12/2025 17:13, Chris Young wrote:
I've had a quick test of that, and getting the same effect (complete system crash as soon as any network access is attempted).
21 Dec 2025 02:56:49 John-Mark Bell <dmarc-noreply@freelists.org>:
On 20/12/2025 20:57, Chris Young wrote:
19 Dec 2025 13:00:33 John-Mark Bell <jmb@netsurf-browser.org>:
> Is it possible to bisect curl releases between 8.11.0 and 8.17.0 to see where it stops working? (the diff between those two versions is nearly 500k lines, so if we can narrow down the search it might be easier to find the problem). I did have to refresh the context of the trivial patch we carry for the m68k-unknown-amigaos build of curl, but I believe it's still applying the same change as before.
>
I'm not sure when I can get to trying that, but certainly seems like a good plan.
In the meantime, I've looked through the configuration differences. Can you please try the Makefile change on jmb/aos3-curl-wat with 8.17.0? That forcibly restores the prior behaviour (which occurred because of a bug in the configure script resulting in it failing to find proto/bsdsocket.h if it isn't on the compiler's default search path). However, if that makes things "work", then something else must already have been wrong (on the basis that this is, afaict, disabling the bsdsocket close/ioctl implementations, which must be working, surely?)
> I'm also assuming that, when you reverted to 8.11.0, it was rebuilt with the updated toolchain and not just an old binary build you had lying around? If not, is it possible to respin 8.11.0 with the new toolchain and see if it still works?
>
Yes, it was the updated toolchain, which I reverted the version and the refreshed patch on.
Great -- thanks for confirming!
J.
No comments:
Post a Comment