From makarius at sketis.net Mon May 4 21:12:07 2026 From: makarius at sketis.net (Makarius) Date: Mon, 4 May 2026 21:12:07 +0200 Subject: Time limiting apply and by commands In-Reply-To: References: Message-ID: <03d7b55b-b796-4032-9113-8695c078a781@sketis.net> On 13/04/2026 12:13, Lawrence Paulson via isabelle-dev wrote: > A year or two ago, I proposed an option to let users impose a time limit on > the execution of proof methods, to make it easier to find performance > regressions after making a change to a large development. It would also make > it easier to find slow steps in any large development. Some people opposed to > this on the grounds that it was “not portable”, which misses the point: it is > not intended to be left in a completed development In the first place. All it > can do is make a proof fail. > It would be a useful feature to have, and extremely easy to implement, so I > hope this time there will be no objections. I am still busy elsewhere, approx. for the next 2 weeks, due to project deadlines. This means, I still did not find time to retrieve the relevant mail threads from 1-2 years ago. Without that it does not make sense to do "extremely easy" things. (Where is the proof for that claim?) We do have fundamental problems with ML timeouts, due to complexities of multithreading and garbage collection. I won't enter that lair right now --- maybe a bit later, when revisiting many PIDE document processing concepts. And that might actually expose better approaches and proper solutions to old problems. I am not following the recent trend to apply adhoc changes blindly, without understanding of the overall context --- of the sources and discussions about approaches to the problem at hand. Makarius From makarius at sketis.net Mon May 4 21:41:45 2026 From: makarius at sketis.net (Makarius) Date: Mon, 4 May 2026 21:41:45 +0200 Subject: NEWS: uniform support for Message_Digest / SHA1 / SHA256 in Isabelle/ML/Scala Message-ID: *** ML *** * Both SHA1 and SHA256 are supported via Isabelle/Scala, with minor changes of the supported operations: Minor INCOMPATIBILITY. SHA1.fake has been replaced by SHA1.parse (including proper checks). Structure Message_Digest provides general operations. The native SHA1 implementation in ML has been discontinued, which may impact performance. *** System *** * Both SHA1 and SHA256 are supported in Isabelle/Scala. SHA1 is still used routinely, but will be eventually phased out. Result is of type Message_Digest.T, with slightly different toString method. Subtle INCOMPATIBILITY: use "rep" instead (as in Isabelle/ML). This refers to Isabelle/50d7fead3758. After dismantling parts of the old "use_thy" loader state, the ML operation SHA1.digest turned out to be unused (in Pure), so that was a good opportunity to dismantle our native SHA1 C-module for Poly/ML, and thus simplify the poly build process and make it more robust. SHA1 remains relevant to some other Isabelle/ML tools (e.g. Sledgehammer, caches), but it now goes through the explicit ML-Scala-function protocol. Adding SHA256 was an easy exercise in re-use of high-quality cryptographic algorithms on the JVM: We may use SHA256 routinely at a later stage, when SHA1 has been generally discredited. There is no immediate need for action, though. My motivation to provide SHA256 right now was a talk at FOSDEM 2026 about the next decade of git. The really interesting talks at that conference were non-git / post-git ones, see also https://fosdem.org/2026/schedule/track/main --- I did not attend this event in person, but a small open source event in Augsburg, where one post-git guy mentioned the very talk that I had watched myself some weeks before. Makarius From lp15 at cam.ac.uk Thu May 7 13:09:45 2026 From: lp15 at cam.ac.uk (Lawrence Paulson) Date: Thu, 7 May 2026 11:09:45 +0000 Subject: The repository is unresponsive Message-ID: hg fetch does not terminate. Larry From makarius at sketis.net Thu May 7 17:24:23 2026 From: makarius at sketis.net (Makarius) Date: Thu, 7 May 2026 17:24:23 +0200 Subject: The repository is unresponsive In-Reply-To: References: Message-ID: <996747a5-2409-4ee8-823d-00d2ec0c5f9d@sketis.net> On 07/05/2026 13:09, Lawrence Paulson via isabelle-dev wrote: > hg fetch does not terminate. It looks fine right now. Makarius