Aw: What is the state of the Windows piping bug ?
Bastian Märkisch <[email protected]>
| Newsgroups | gmane.comp.graphics.gnuplot.devel |
|---|---|
| Message-ID | <trinity-2d7e7a5f-8c41-4f6a-8dd5-e2456c8facae-1626332661395@3c-app-webde-bap44> |
Ethan, The Windows builds for 5.4.2 are just that - builds from the 5.4.2 source release. The piping bug #2412 has been confirmed fixed by the commit you mentioned. It was introduced by an attempt to fix bug #2204 and I am waiting for feedback on this one. We should only release 5.4.3 with this issue confirmed, too. In any case the Windows binary already now resolves some serious platform-specific bugs, including a crash due to initialization errors, see #2223, #2227, #2286, #2402. So the advisory on not using it at all is too strong in my opinion. Btw. please note that the "-os2" file is for OS/2, eComStation, ArcaOS and "_dj" is a DOS (32bit) version. Bastian Gesendet: Donnerstag, 15. Juli 2021 um 06:30 Uhr Von: "Ethan A Merritt" <[email protected]> An: [email protected] Cc: "Bastian Märkisch" <[email protected]> Betreff: What is the state of the Windows piping bug ? Bastian: Your fix for the WIndows piping bug went into the stable branch on 10 July. This was after the nominal release date for version 5.4.2 The Windows binaries on SourceForge are timestamped that same day. gp542-os2-emx.zip 2021-07-11 3.4 MB gp541-win64-mingw.7z 2021-07-10 28.8 MB gp542_dj.zip 2021-07-10 4.0 MB gp542-win64-mingw.exe 2021-07-10 31.7 MB Is the fix included in these binaries or not? If it is then the README on the download site should say so (right now it advises not to use the Windows binaries) If it is not - should we put out a quick version 5.4.3 for Windows? Commit a501a67a Author: Bastian Maerkisch <[email protected]> Date: Sat Jul 10 20:41:40 2021 +0200 Fix handling Window messages for input from a pipe As pointed out in bug report #2412, MsgWaitForMultipleObjects() does not work as expected if the input comes from a pipe. This reverts 651af6267b "Wait for pipe events without additional thread", which aimed at speeding-up piped input, Instead of recreating the input thread for every single character, we now keep the thread alive and signal an event object on new input. This should still be much faster. Bugs #2412 and #2204 _______________________________________________ gnuplot-beta mailing list [email protected] Membership management via: https://lists.sourceforge.net/lists/listinfo/gnuplot-beta