Bug#392942: Regression in selection lists, probably due to #384787
Hello Alastair, AFAICT Eugeniy's analysis shows that this issue is unrelated to his patch for #384787, but rather a separate slang issue. What is your take on it and how can we best proceed to resolve it? If at all possible, I'd like to see it resolved before the RC1 release, but that is going to be very tight. If not, we could delay it until the next release candidate. Eugeniy: are you by any chance still working on this? Thanks, FJP pgpoOyajnEcbo.pgp Description: PGP signature
Bug#392942: Regression in selection lists, probably due to #384787
Frans Pop wrote: Hello Alastair, AFAICT Eugeniy's analysis shows that this issue is unrelated to his patch for #384787, but rather a separate slang issue. What is your take on it and how can we best proceed to resolve it? If at all possible, I'd like to see it resolved before the RC1 release, but that is going to be very tight. If not, we could delay it until the next release candidate. Eugeniy: are you by any chance still working on this? Thanks, FJP I've been looking at this bug last night and today (a bit more awake). The problem is with a patch 080_wide_chars that broke jed (see #369152) and that I removed; I think this is patch needs fixing rather than removing. The error is in SLutf8_skip_chars() and I hope to work on it tonight if I get a chance. Patches and well-tested NMUs welcome. Regards Alastair -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]
Bug#392942: Regression in selection lists, probably due to #384787
18 жовтня 2006 о 16:33 +0200 Frans Pop написав(-ла): Eugeniy: are you by any chance still working on this? No, I am not going to work on that piece of crap called slang2. I was able to read code of slang1 but not slang2. -- Eugeniy Meshcheryakov signature.asc Description: Digital signature
Bug#392942: Regression in selection lists, probably due to #384787
Package: newt Version: 0.52.2-8 Severity: serious I just noticed an important issue in newt that shows up in daily builds of the installer. For languages (CJK) using double width characters, calculation of line length is incorrect. It only affects a few languages, but unfortunately affects them in a bad way. These screenshots show the issue: http://people.debian.org/~fjp/d-i/d-i_newt_regression.png http://people.debian.org/~fjp/d-i/d-i_newt_regression2.png I'm very sorry that I missed this when testing the fixed input box handling (which really is fixed). Eugeniy: do you maybe have time to look into this? Cheers, FJP pgpOMKMiwPn1A.pgp Description: PGP signature
Bug#392942: Regression in selection lists, probably due to #384787
This bug is really related to #384787, in sence that something was change in newt, or more likely - in slang. It looks like functions SLsmg_write_string and SLsmg_write_nstring now work incorrecly. This bug can be seen without d-i by using the following scripts: whiptail --menu test 20 50 3 1 日本語 2 English 3 Українська\ мова And for bug similar to #384787 but with CJK: whiptail --inputbox test 20 50 日本語 14 жовтня 2006 о 13:49 +0200 Frans Pop написав(-ла): I just noticed an important issue in newt that shows up in daily builds of the installer. For languages (CJK) using double width characters, calculation of line length is incorrect. It only affects a few languages, but unfortunately affects them in a bad way. These screenshots show the issue: http://people.debian.org/~fjp/d-i/d-i_newt_regression.png http://people.debian.org/~fjp/d-i/d-i_newt_regression2.png I'm very sorry that I missed this when testing the fixed input box handling (which really is fixed). -- Eugeniy Meshcheryakov signature.asc Description: Digital signature
Bug#392942: Regression in selection lists, probably due to #384787
clone 392942 -1 reassing -1 slang2 block 392942 by -1 thanks This is cleany bug in slang2. slsmg.c contains: void SLsmg_write_nstring (char *str, unsigned int n) { ... if (UTF8_Mode) umax = SLutf8_skip_chars (u, u+width, n, width, 0); ... SLsmg_write_chars (u, umax); } while (width++ n) SLsmg_write_chars (blank, blank+1); } from last line: 'n' is string width. Let's look at SLutf8_skip_chars, in slutf8.c: SLuchar_Type *SLutf8_skip_chars (SLuchar_Type *s, SLuchar_Type *smax, unsigned int num, unsigned int *dnum, int ignore_combining) { unsigned int n; n = 0; while ((n num) (s smax)) { ... /* here code that modifies n always adds 1, nothing depends on * character width */ ... So here num, aka n in previous function, is used as number of characters and not as string width. /me dreams about time when slang's upstrem mainainers enlighten themself by existance of wc* and mb* functions. 14 жовтня 2006 о 14:28 +0200 eugen написав(-ла): This bug is really related to #384787, in sence that something was change in newt, or more likely - in slang. It looks like functions SLsmg_write_string and SLsmg_write_nstring now work incorrecly. This bug can be seen without d-i by using the following scripts: whiptail --menu test 20 50 3 1 日本語 2 English 3 Українська\ мова And for bug similar to #384787 but with CJK: whiptail --inputbox test 20 50 日本語 14 жовтня 2006 о 13:49 +0200 Frans Pop написав(-ла): I just noticed an important issue in newt that shows up in daily builds of the installer. For languages (CJK) using double width characters, calculation of line length is incorrect. It only affects a few languages, but unfortunately affects them in a bad way. These screenshots show the issue: http://people.debian.org/~fjp/d-i/d-i_newt_regression.png http://people.debian.org/~fjp/d-i/d-i_newt_regression2.png I'm very sorry that I missed this when testing the fixed input box handling (which really is fixed). -- Eugeniy Meshcheryakov -- Eugeniy Meshcheryakov signature.asc Description: Digital signature