| Message ID | 20200713184751.23534-1-asierra@xes-inc.com |
|---|---|
| Headers | show |
| Series | Use bzip2 for X11 PFC font compression | expand |
Hello Aaron, Thanks for this work! On Mon, 13 Jul 2020 13:47:49 -0500 Aaron Sierra <asierra@xes-inc.com> wrote: > Gzip has been used as the default compressor for PCF fonts, but this > series changes the default compressor to bzip2 for a few reasons: > > 1. Even with the latest gzip, these seemingly synonymous pipelines > produce different output, but this issue does not exist with bzip2: > > $ cat /path/to/file | gzip > /path/to/file.gz > $ gzip < /path/to/file > /path/to/file.gz > > 2. Prior to gzip 1.10, the compression pipeline used with PCF fonts was > not reproducible due to the implicit -N/--name injecting a timestamp: > > * cat /path/to/file | gzip > /path/to/file.gz > > 3. The BR2_USE_WCHAR dependency of the gzip package tarnishes the appeal > of using host-gzip to provide reproducible output. This argument seems pretty weird. The fact that gzip needs BR2_USE_WCHAR on the target doesn't at all prevent from building host-gzip. We have plenty of host packages that need wchar packages, and we simply assume the host system as wide char support available. So this third argument is a bit "moot", especially since host-xapp-mkfontscale already has a dependency on host-gzip, which builds a gzip 1.10, so it shouldn't be affected by the problem you describe. So that leaves us with just argument (1), correct ? Thomas
On Mon, 13 Jul 2020 21:32:25 +0200 Thomas Petazzoni <thomas.petazzoni@bootlin.com> wrote: > So this third argument is a bit "moot", especially since > host-xapp-mkfontscale already has a dependency on host-gzip, which Sorry, host-xapp-mkfontscale has a dependency on host-zlib, but adding a dependency on host-gzip should not cause any problem. Thomas
----- Original Message ----- > From: "Thomas Petazzoni" <thomas.petazzoni@bootlin.com> > Sent: Monday, July 13, 2020 2:32:25 PM Hi Thomas, Thanks for your review. > Hello Aaron, > > Thanks for this work! > > On Mon, 13 Jul 2020 13:47:49 -0500 > Aaron Sierra <asierra@xes-inc.com> wrote: > >> Gzip has been used as the default compressor for PCF fonts, but this >> series changes the default compressor to bzip2 for a few reasons: >> >> 1. Even with the latest gzip, these seemingly synonymous pipelines >> produce different output, but this issue does not exist with bzip2: >> >> $ cat /path/to/file | gzip > /path/to/file.gz >> $ gzip < /path/to/file > /path/to/file.gz >> >> 2. Prior to gzip 1.10, the compression pipeline used with PCF fonts was >> not reproducible due to the implicit -N/--name injecting a timestamp: >> >> * cat /path/to/file | gzip > /path/to/file.gz >> >> 3. The BR2_USE_WCHAR dependency of the gzip package tarnishes the appeal >> of using host-gzip to provide reproducible output. > > This argument seems pretty weird. The fact that gzip needs > BR2_USE_WCHAR on the target doesn't at all prevent from building > host-gzip. We have plenty of host packages that need wchar packages, > and we simply assume the host system as wide char support available. > > So this third argument is a bit "moot", especially since > host-xapp-mkfontscale already has a dependency on host-gzip, which > builds a gzip 1.10, so it shouldn't be affected by the problem you > describe. OK, then maybe adding a host-gzip dependency would be the better solution. I've found some evidence that this patchset isn't complete with respect to X itself using bzip-compressed fonts :( > So that leaves us with just argument (1), correct ? Well, I think that (2) or (3) would be needed as the real justification for switching compression mechanisms. Let me investigate the host-gzip path now that I know that isn't a compatibility problem. -Aaron > Thomas > -- > Thomas Petazzoni, CTO, Bootlin > Embedded Linux and Kernel engineering > https://bootlin.com
On Mon, 13 Jul 2020 14:55:10 -0500 (CDT) Aaron Sierra <asierra@xes-inc.com> wrote: > >> 1. Even with the latest gzip, these seemingly synonymous pipelines > >> produce different output, but this issue does not exist with bzip2: > >> > >> $ cat /path/to/file | gzip > /path/to/file.gz > >> $ gzip < /path/to/file > /path/to/file.gz > >> > >> 2. Prior to gzip 1.10, the compression pipeline used with PCF fonts was > >> not reproducible due to the implicit -N/--name injecting a timestamp: > >> > >> * cat /path/to/file | gzip > /path/to/file.gz > >> > >> 3. The BR2_USE_WCHAR dependency of the gzip package tarnishes the appeal > >> of using host-gzip to provide reproducible output. > > > > This argument seems pretty weird. The fact that gzip needs > > BR2_USE_WCHAR on the target doesn't at all prevent from building > > host-gzip. We have plenty of host packages that need wchar packages, > > and we simply assume the host system as wide char support available. > > > > So this third argument is a bit "moot", especially since > > host-xapp-mkfontscale already has a dependency on host-gzip, which > > builds a gzip 1.10, so it shouldn't be affected by the problem you > > describe. > > OK, then maybe adding a host-gzip dependency would be the better solution. > I've found some evidence that this patchset isn't complete with respect to > X itself using bzip-compressed fonts :( > > > So that leaves us with just argument (1), correct ? > > Well, I think that (2) or (3) would be needed as the real justification for > switching compression mechanisms. Let me investigate the host-gzip path now > that I know that isn't a compatibility problem. OK, thanks! Thomas